Datasets:

babytreecc

/

ICLR2025review

like 7

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- Is it possible to implement a random-policy baseline where you randomly chose a set of (naturally collected) datapoints from a data pool? The no-state baseline has flavor of this baseline but LLM-informed decisions could be biased. \n- Is it possible to compare this approach with active learning, in which instead of doing data generation, you do data *selection* and ask models to generate only synthetic labels, but not synthetic inputs?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- Tackle a timely and interesting problem. \n- Provide the necessary infrastructure for the community to study the problem, opening up opportunities for future contributions. \n- Consider various data generation strategies,\n- Well-desgined experiments which demonstrate the effectiveness of the proposed approaches and conduct insightful analyses." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces Gym environments for data synthesis, framing the problem as sequential decision-making. In these environments, actions correspond to data-generation plans, and states represent the performance summary of a student model. The paper implements environments for three tasks: visual question answering (VQA), math, and code generation. Each environment offers three state representations: open-ended, skill-list, and skill-tree. Additionally, it proposes an LLM-based policy for data generation. Experimental results demonstrate that the LLM can make strategically effective choices based on environment-state information." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* The paper is currently dense and difficult to follow. The introduction includes excessive implementation details, which detract from providing a simple, high-level intuition. Using a specific task example to guide readers through the core concepts would make the paper more accessible.\n\n* The paper focuses solely on the data generation plan rather than a full, end-to-end data generation process. It relies on a fixed, off-the-shelf data-generation engine that cannot be modified. The authors should admit this limitation and discuss potential strategies for overcoming it.\n\n* The quality of the data-generation engine can impact both student performance and the data-generation plan itself. Current approaches do not take into account the data-generation engine capabilities in the design of the policy or the evaluation of the student. For instance, poor student performance might result from the engine producing low-quality data on a specific skill, which could prompt the policy to avoid querying the engine for that skill.\n\n* The learning procedure can be resource-intensive. The authors should report the time, cost, and computing resources used for the experiments." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "N/A" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "In the Experiments section, the authors mention that the baseline student model should not have been heavily post-trained so that there are rooms for further improvements. However, it would be beneficial to provide additional evidence and details to support the claim that the student's performance is improved due to the added data points rather than insufficient training. For instance, the training protocol involved a fixed 10-epoch training period; it remains unclear whether the model had reached convergence within this timeframe or if the introduction of new data points accelerated convergence. Further clarification on this aspect would enhance the overall validity of the results.\n\nAlso the result would be more sound if more quantitative and qualitative results for skill discovery is reported in this paper." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper presents a novel and insightful perspective on the autonomous data generation problem, leveraging principles from reinforcement learning to conceptualize it as a sequential decision-making process. The authors provide a thorough explanation of this approach, the motivations behind and the underlying mechanics.\n\nThis paper proposed a modular framework/testbed that can be easily adapted to various tasks, showcasing its versatility and potential for widespread applicability. The authors demonstrate the effectiveness of their approach through experiments on 3 tasks of multiple modalities, including text, image, and code generation, yielding promising early results." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a modular system for automated data generation, designed to minimize the need for human annotations. The proposed approach employs a reinforcement learning-inspired methodology, decomposing the process into a sequence of action predictions (data generation policy) based on state information (feedback from model errors) in an iterative manner. The effectiveness of this approach is demonstrated through three diverse tasks, encompassing text, image, and code generation across different modalities." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The experiment part should be conducted more thoroughly: specifically, creating a test set that incorporates newly generated data points from the data generation agent and reporting evaluation results for each retrained model over successive iterations would provide more comprehensive insights into the system's performance." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- How does the performance of the data generation agents change over longer iterations? The paper truncates experiments when performance increases, but it would be insightful to explore whether performance plateaus or continuously increase over extended training.\n- Is the total training data allocation fixed in each environment, or does it vary dynamically? The methodology mentions rebalancing but lacks clarity on how these allocations adjust adaptively based on feedback." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- Good contribution to automated data generation for model improvement.\n- Clearly written with structured sections explaining each environment type and experimental results." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents DataEnvGym, a framework designed to simulate environments for data generation agents. These agents iteratively generate synthetic data to address weaknesses in student models, aiming to improve model performance across tasks like mathematics, programming, and visual question answering. DataEnvGym provides various structured environments (Open-Ended, Skill-List, and Skill-Tree) where data generation agents create targeted training examples based on feedback from the student model, offering a dynamic approach to automated model improvement." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The paper should clarify early on that the focus is on synthetic data generation for training purposes, as this underpins the motivation for the approach.\n- Important related works on algorithms using feedback from training to generate the next training environments are missing [1, 2, 3, 4].\n- Lines 460 - 465, I believe there is a typo whereby it says that “each experiment is truncated once the performance consistently decreases for multiple iterations”. Should it be “increases”?\n- Repeated runs of experiments without confidence intervals will be valuable, especially since the variance of performance seems to be very high.\n\n[1] Sudhakaran, S., González-Duque, M., Freiberger, M., Glanois, C., Najarro, E., & Risi, S. (2024). Mariogpt: Open-ended text2level generation through large language models. Advances in Neural Information Processing Systems, 36.\n[2] Todd, G., Earle, S., Nasir, M. U., Green, M. C., & Togelius, J. (2023, April). Level generation through large language models. In Proceedings of the 18th International Conference on the Foundations of Digital Games (pp. 1-8).\n[3] Zhang, J., Lehman, J., Stanley, K., & Clune, J. (2023). Omni: Open-endedness via models of human notions of interestingness. arXiv preprint arXiv:2306.01711.\n[4] Faldor, M., Zhang, J., Cully, A., & Clune, J. (2024). OMNI-EPIC: Open-endedness via Models of human Notions of Interestingness with Environments Programmed in Code. arXiv preprint arXiv:2405.15568." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "Yes, Discrimination / bias / fairness concerns", "Yes, Potentially harmful insights, methodologies and applications" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Limited Evaluation of Agent Architectures: The paper primarily focuses on introducing the DataEnvGym environment, but the evaluation of data generation agents is limited to relatively simple baseline policies. Exploring more sophisticated agent architectures, such as reinforcement learning agents (e.g., using policy gradient methods, Q-learning) or agents incorporating larger language models for planning and decision-making (similar to the approaches used in Shimabucoro et al. (2024), would substantially strengthen the paper. A systematic comparison of different agent architectures in terms of their effectiveness in improving student models, their sample efficiency, and their computational cost would provide valuable insights and contribute to a better understanding of the challenges and opportunities in automated data generation.\n\nLimited Analysis of Skill Discovery Quality: The paper briefly discusses the impact of oracle skills on student performance but doesn't delve deeply into the quality of the skills discovered by the proposed LLM-based method. A more thorough analysis is needed to understand the strengths and limitations of the skill discovery module. This could involve quantitative measures of skill quality, such as measuring their coherence, coverage, and relevance to the target task, or qualitative analysis by human experts. Investigating how the quality of the discovered skills affects the performance of the data generation agents and the resulting student models would strengthen the paper's contribution. Exploring alternative skill discovery methods (e.g., clustering-based approaches, topic modeling) and comparing their effectiveness with the proposed method would further enhance the analysis.\n\nLack of Comparison with Existing Methods: The paper positions DataEnvGym as a novel approach for model improvement, but it lacks a direct comparison with existing methods like curriculum learning (Bengio et al., 2009) or active learning (Settles, 2009). Evaluating how DataEnvGym compares to these established techniques in terms of student model performance, data efficiency, and computational cost would provide valuable context and highlight the advantages of the proposed framework. This would also clarify the specific niche and contribution of DataEnvGym within the broader landscape of model improvement techniques.\n\nLimited Discussion of Scalability: The experiments in the paper are conducted with relatively small datasets and models. It's essential to address the scalability of DataEnvGym to more realistic scenarios involving larger datasets, more complex models, and a broader range of skills. Discussing the computational challenges and potential optimizations for scaling the framework to more demanding settings would strengthen the paper's practical relevance. For instance, how can the computational cost of LLM-based data generation be reduced while maintaining data quality? How can the skill discovery and agent training processes be optimized for larger datasets? Addressing these questions would provide valuable insights for future research and practical applications." }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "Novel Problem: Automating data generation to improve models is a significant challenge with practical applications. This work directly addresses this problem with a novel approach.\n\nWell-Defined Framework: DataEnvGym is presented as a well-defined framework with clear components (trainer, evaluator, data generation policy, data generation engine) and different levels of structure (open-ended, skill-list, skill-tree). This structure makes the problem tractable and facilitates modular development and testing.\n\nMultiple Tasks and Domains: The inclusion of experiments across diverse tasks (mathematics, programming, visual question answering) and with different student models demonstrates the generalizability of the framework.\n\nPromising Results: The initial results showing improved student model performance across tasks and environments are encouraging and suggest the potential of this approach. The analysis of difficulty/rarity and training dynamics adds value.\n\nOpen-Source Release: The commitment to publicly releasing the code and leaderboard promotes reproducibility and encourages further research in this area." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces DataEnvGym, a novel testbed of teacher environments for developing data generation agents that iteratively improve student models by generating targeted training data. DataEnvGym frames data generation as a sequential decision-making task where an agent, comprising a data generation policy and engine, interacts with an environment that provides feedback from a student model. The agent's goal is to improve student model performance by generating training data based on student feedback (errors or weak skills). DataEnvGym offers multiple instantiations of teacher environments across three levels of structure: open-ended, skill-list, and skill-tree, each with varying levels of scaffolding support. Experiments across text and image-based tasks (mathematics, programming, and visual question answering) demonstrate that example agents within DataEnvGym can iteratively improve student model performance. Furthermore, the authors analyze the impact of state information, environment structure, and skill discovery quality on agent performance and student learning. The paper concludes that DataEnvGym, with its modular design and support for diverse tasks and student models, provides a valuable platform for developing and evaluating data generation agents, engines, and feedback mechanisms for automated model improvement. The code and leaderboard will be publicly released." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Limited Evaluation of Agent Architectures: The focus is primarily on the environment itself, with less emphasis on the architecture and training of the data generation agents. While baseline agents are provided, more sophisticated agent designs (e.g., reinforcement learning agents, agents leveraging larger language models) and their systematic evaluation would significantly strengthen the paper. How do different agent architectures compare in terms of effectiveness and efficiency? Are there specific architectural choices that are particularly well-suited for this task?\n\nOver-Reliance on LLMs for Data Generation: While using LLMs for data generation is a reasonable starting point, it raises concerns about the quality and diversity of the generated data. Exploring alternative data generation methods (e.g., data augmentation techniques, programmatic data generation) and comparing their effectiveness with LLM-based generation would be valuable. How robust is the framework to the quality of the generated data?\n\nLimited Analysis of Skill Discovery Quality: While the paper briefly touches upon the impact of skill discovery quality, a more thorough investigation is needed. How does the quality of the discovered skills affect the performance of the data generation agents and the student models? What are the limitations of the current skill discovery method, and how can it be improved? Quantitative analysis of skill quality (e.g., measuring coherence, coverage, and relevance) would strengthen the paper.\n\nLack of Comparison with Existing Methods: While related work on knowledge distillation and model weakness discovery is discussed, there's no direct comparison with existing methods for model improvement. How does DataEnvGym compare to techniques like curriculum learning or active learning in terms of effectiveness and efficiency? Including such comparisons would better contextualize the contributions and highlight the advantages of the proposed approach.\n\nLimited Discussion of Scalability: The experiments are conducted with relatively small datasets and models. How does DataEnvGym scale to larger datasets and more complex models? What are the computational challenges associated with training data generation agents in more realistic settings? Addressing these scalability concerns is crucial for practical applications." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024dataenvgym,\ntitle={DataEnvGym: Data Generation Agents in Teacher Environments with Student Feedback},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=00SnKBGTsz},\nnote={under review}\n}" }, "abstract": { "value": "The process of creating training data to teach models is currently driven by humans, who manually analyze model weaknesses and plan how to create data that improves a student model. Recent approaches using large language models (LLMs) as annotators reduce human annotation effort, but still require humans to interpret feedback from evaluations and control the LLM to produce data the student needs. Automating this labor-intensive process by creating autonomous data generation agents – or teachers – is desirable, but requires environments that can simulate the feedback-driven, iterative, closed loop of data creation. To enable rapid and scalable testing for such agents and their modules, we introduce DataEnvGym, a testbed of teacher environments for data generation agents. DataEnvGym frames data generation as a sequential decision-making task, involving an agent consisting of a data generation policy (which generates a plan for creating training data) and a data generation engine (which transforms the plan into data), inside an environment that provides feedback from a student. The agent’s end goal is to improve student model performance. Students are iteratively trained and evaluated on generated data, with their feedback (in the form of errors or weak skills) being reported to the agent after each iteration. As a general-purpose testbed, DataEnvGym includes multiple instantiations of teacher environments across three levels of structure in the state representation and action space, with varying levels of scaffolding support. More structured environments are based on automatically-inferred skills and offer a higher degree of interpretability and control over the curriculum. We support developing and testing data generation agents in three diverse tasks covering both text and images (mathematics, programming, and visual question answering) and test multiple student models. We find that example agents in our teaching environments can iteratively improve students across diverse tasks and settings. Moreover, we show that environments can teach different skill levels and can be used to test variants of key modules, pointing to directions of future work in improving data generation agents, engines, and feedback mechanisms. We will publicly release our code and leaderboard." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "iterative data generation", "llm agent", "lifelong learning" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/2fabe224ce80b58518b3e21579a58af4d807e6d7.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "DataEnvGym: Data Generation Agents in Teacher Environments with Student Feedback" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The authors need to provide further experiments and analyses to demonstrate the reliability of the questions generated by this method, such as incorporating the performance of human experts or introducing relevant methods for quality control of the questions in the methods section.\n\n2. Also, more analysis of the evaluation results should be included. For example, what are the main types of errors introduced by attacks across different turns? Which specific diseases or problem types is the target LLM less robust against? By supplementing these analyses, further insights can be provided for the development of medical LLMs." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "+ This paper examines the robustness of LLMs in the clinical decision-making process, a critical aspect of their application in the medical domain.\n\n+ The evaluation results demonstrate that current LLMs lack robustness in the clinical decision-making process, offering valuable insights for the development of medical LLMs." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper investigates the robustness of large language models in handling medical QA tasks by introducing a new evaluation method, MedFuzz. For each multiple-choice question in the original benchmarks, MedFuzz uses an LLM (referred to as the attacker LLM) to reformulate questions by adding patient characteristics that may introduce social bias without affecting the clinical decision-making process. If the target LLM answers correctly, the attacker LLM is prompted to generate additional distracting questions based on the target LLM’s feedback. Additionally, a non-parametric statistical significance test was developed by prompting the attacker LLM to create questions with patient characteristics that avoid social bias. Using this evaluation method, the authors tested seven LLMs and found a significant performance drop across all models. Moreover, they observed that when current LLMs answer incorrectly, they tend not to reference the added biased information, indicating inconsistency in faithfully adhering to the clinical decision-making process." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "+ A major weakness of this paper is the faithfulness of the reformulated questions. The proposed MedFuzz method relies solely on prompt engineering with the attacker LLM (GPT-4) to modify original MedQA questions, making the attack process difficult to control. The attacker LLM may potentially alter critical information in the original questions, resulting in less reliable reformulated questions. The example in Section 3.1 also demonstrates that the attacker LLM added extensive information about the patient’s family medical history, consultation history, and medication history. These details are highly relevant in real clinical diagnosis and can significantly influence a doctor’s assessment of the patient’s condition.\n\n+ Moreover, although the authors propose a non-parametric statistical significance test, they do not provide the full distribution of p-values across the MedQA benchmark. In line 485, they note that for the successful attacks they selected, the p-values are <1/30, 0.1, 0.16, 0.5, and 0.63. Here, the p-value represents the probability that a control fuzz is more challenging than the original fuzz. Therefore, cases with p-values of 0.5 and 0.63 suggest that the performance decline in the target LLM is due to the perturbations themselves, rather than social bias.\n\n+ For the study of target LLM's faithfulness, it is important to also study the proportion of CoT that mentions the critical information in the original MedQA benchmark for comparison with the results provided in Figure 2B. Additionally, the authors should provide more information to help readers understand the specific process of this study. For example, how many cases were analyzed? Was the determination of whether fuzzed information was included made manually, or was an automated algorithm used?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "NA. Authors have provided an ethics statement in the draft as well." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "•\tThe authors can clarify how their approach to adversarial attacks differs from the misinformation approach in [1].\n\n•\tThe authors can clarify why unfaithfulness of generated responses is a crucial dimension to consider.\n\n•\tSection 2.2 Lines 104: The authors mention “two ways” in which MedFuzz differs from other adversarial ML approaches, though only one distinction is clear in the draft. I’m assuming the second way is the use of semantically coherent changes to the text. These few lines can probably be rephrased to add clarity.\n\n•\tThe authors have conducted their experiments on the MedQA dataset and taken advantage of a constraint imposed in the curation of this dataset. The authors could potentially add broad guidelines to expand on the fuzzing idea for other medical datasets. \n\n•\tHow can the authors ensure that the GPT-4 generated attack retains the same answer as the original QA pair being perturbed? Is there a possibility to evaluate this with the help of domain experts?\n\n•\tHow is the value of K set in Algorithm 1? This can be elaborated on in the Appendix section.\n\n•\tDoes the finding that LLM CoT does not mention the fuzzed information provide a way forward to identify adversarial inputs?\n\n•\tAnother interesting avenue would be to examine how different kinds of LLMs perform when used as the attacking/ target LLM. For example, can a smaller model generate adversarial inputs faster than a larger model like GPT-4?\n\n•\tMinor Comment: Is line 10 a duplicate of line 11 in Algorithm 1?\n\n[1] Han T, Nebelung S, Khader F, Wang T, Müller-Franzes G, Kuhl C, Försch S, Kleesiek J, Haarburger C, Bressem KK, Kather JN. Medical large language models are susceptible to targeted misinformation attacks. npj Digital Medicine. 2024 Oct 23;7(1):288." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "•\tClarity: The paper is well written and easy to follow along. The authors have given adequate and clear examples at appropriate locations in the draft to aid readability. Good use of illustrations after consultation with domain experts (clinical collaborators in this case). The authors have also acknowledged the limitation of using contaminated training data.\n\n•\tOriginality: The idea to use social biases a clever way to incorporate real life information into the MedQA dataset.\n\n•\tQuality: The evaluation involves the use of proprietary vs open source and general purpose vs domain specific models. The experiment settings for reproducibility like temperature have been provided. The approach should be easy enough to reproduce. \n\n•\tSignificance: The authors have tackled a relevant problem that needs to be addressed, given the rapid pace of the domain." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes an automated red teaming approach to attack LLMs. They attempt this in the medical context by modifying medical Q&A datasets (specifically on MedQA), by violating assumptions that do not hold good in real life settings. The goal of MedFuzz is to make LLMs provide the wrong answer while ensuring that clinicians can still provide the right answer. The authors have identified a crucial problem with the evaluations of LLMs in the medical domain and provided a way to generate a more realistic dataset to aid subsequent LLM evaluation. The novelty lies in the proposed dataset from MedFuzz and the statistical evaluation used to check if the attack was successful." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "•\tIn the case of MedQA dataset, the authors have identified a social bias which may be present in real life situations, which are removed in the original benchmark. It is unclear how easy it is to identify and exploit such peculiarities in other medical benchmarking datasets like MedMCQA[1], PubMedQA[2] etc.\n\n•\tThe authors create the adversarial questions by an iterative multi-turn approach. Although the authors allude to the target LLM forgetting about previous Q&A attempts, would the approach be better validated if the evaluation is done in a single-turn manner?\n\n•\tThe authors, in step 4, only validate the statistical significance of 4 individual interesting cases. How would this change if considered for all successful cases?\n\n[1] Pal A, Umapathi LK, Sankarasubbu M. Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering. InConference on health, inference, and learning 2022 Apr 6 (pp. 248-260). PMLR.\n\n[2] Jin Q, Dhingra B, Liu Z, Cohen WW, Lu X. Pubmedqa: A dataset for biomedical research question answering. arXiv preprint arXiv:1909.06146. 2019 Sep 13." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* Why was MedQA the only dataset used? There are a few other multiple choice medical QA ones liked MedMCQA, PubMedQA, and MMLU Clinical topics. Why MedQA?\n* Why was only GPT-4 used as the attacker LLM? Seemingly there are other open source ones that have just as much medical knowledge especially looking at the fine-tuned example. \n* The workflow for the Step 2 is quite a few iterative turns. Are they all necessary to generate grounded ones? Is this workflow generalizable to other LLMs? Or is it GPT-4 specific?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The idea of the paper is interesting -- existing medical QA datasets are fairly simplified and may not appropriately represent real-world clinical settings. Thus, there is a need to understand how safe LLM usage is for the medical domain via robustness analysis.\n* The intuition for the adversarial biasing comes from medical domain understanding of the benchmark constructions.\n* Authors benchmark 3 closed LLMS and 4 open-source, medically fine-tuned LLMs." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes an adversarial method for evaluating LLM performance on medical question-answering benchmarks to assess their robustness in real-world clinical settings. The idea is to automatically generate new question-answer pairs from the existing benchmark such that they still represent realistic scenarios (e.g., including additional patient information) but the answers remain the same. The experiment results demonstrate that various baseline LLMs can be tricked into providing incorrect answers." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* One of the major claims of the method is that it will generate new questions that are semantically coherent and will not fool clinicians. However, there is no empirical proof that this is the case other than the analysis of a handful of case studies (one is presented in the main text). The prompt contains instructions for the attacker LLM it should not change the default answer but GPT-4 is not always guarenteed to follow the instructions or have all the correct medical knowledge appropriate.\n* Is there a reason why general domain adversarial prompting wasn't shown to be sufficient? A few studies are listed in 2.2 (first sentence) but no preliminary studies or experimental studies are shown to support this.\n* GPT-4 is chosen as the attacker LLM, but the question is why aren't other open-source models explored? In looking at OpenBIOLLM-70B performance, this also looks like a reasonable comparison to try and might even generate harder cases with less of the computation cost.\n* One of the comments in the introduction was the that existing benchmarks are not challenging enough including reducing real-life clinical situations to canonical multiple choice questions. Is there a reason why only one dataset was included and it was a multiple-choice one?\n* The statistical test is proposed to identify the significance of a successful attack using control fuzzes and to select the case studies, but what about the general distribution for the MedQA dataset? How stable is it broadly in identifying how significant a successful attack is? I understand this can be computationally intensive and costly but that also raises a bit of questions regarding the applicability of the method if it can't be done at scale. \n* The presentation could have been improved to provide some intuition at the beginning with potentially a simpler case study where less was added to make the LLM response change. Similarly, some of the text is written in a less digestible format. For example, the introduction of the test statistic could be improved by introducing notation first and then how you might compute it to understand what the statistic is looking to capture.\n* The citation format is incorrect, please use \\citep instead of \\cite as it detracts from readability." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "In the MedFuzz study, patient characteristics (PC) such as age, gender, race, and socioeconomic factors are added as perturbations to induce confusion in LLMs. One specific example presented by the authors is the use of “excessive hospital service usage by low-income patients.” This type of information could inadvertently reinforce social biases or perpetuate negative perceptions about certain demographic groups, rather than reflect clinical validity or fairness.\n\nWhen such characteristics are introduced as confusion-inducing factors, there is a risk that essential background information—critical for accurate diagnosis and treatment—could lead to biased outcomes. Therefore, further clarification and evaluation are needed to ensure that MedFuzz’s inclusion of such data as perturbations aligns with clinical relevance and fairness, and to mitigate any potential reinforcement of harmful social biases in the model." }, "flag_for_ethics_review": { "value": [ "Yes, Discrimination / bias / fairness concerns" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. It would be helpful to have specific examples illustrating the risks posed by the simplified assumptions in traditional benchmarks within clinical settings. For instance, if omitting certain patient characteristics or clinical contexts could lead to diagnostic errors, providing these examples would clarify the importance of this study for readers and highlight its relevance.\n\n2. I am curious whether the patient characteristics (e.g., age, gender) and social bias information added as perturbations in MedFuzz genuinely act as confusion factors within actual clinical environments. These details often serve as crucial data points in clinical decision-making, so further explanation on how these elements were deemed appropriate as confusion-inducing factors would enhance the clinical validity of this study.\n\n3. A clear explanation regarding the rationale for setting the perturbation iteration count to K=5 would be beneficial. For instance, do you have experimental results comparing the initial attack (K=1) with subsequent attacks (K=5) to illustrate how the LLM maintains performance with increasing perturbation levels? Such a comparison could provide a more reliable basis for evaluating the impact of iteration count on robustness in this study." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. This paper introduces MedFuzz, a novel approach for testing the robustness of large language models (LLMs) in clinical contexts, which addresses the simplifications found in traditional benchmarks. MedFuzz is distinct in its approach by adding specific patient characteristics and social bias information to simulate the complexity of real-world clinical scenarios. This innovative framework offers a new direction for assessing LLM robustness by examining potential vulnerabilities in medical question-answering settings.\n\n2. The paper clearly explains the concept of MedFuzz and its application, particularly in using patient characteristics and bias elements to test model robustness. The experimental procedures and components are consistently described, making the study's objectives and methodology easy for readers to follow.\n\n3. MedFuzz presents a significant contribution as it provides a framework to evaluate how LLMs may perform in real clinical settings, beyond simplified benchmarks. This work has high practical relevance for the safe implementation of LLMs in healthcare by strengthening robustness assessment and reducing potential errors. It contributes an essential tool for enhancing LLM applicability in clinical practice, highlighting the importance of robustness in medical AI." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes MedFuzz, a novel approach designed to evaluate the robustness of large language models (LLMs) in medical question-answering contexts. MedFuzz introduces controlled perturbations in input text by adding patient characteristics (PC) and social bias information to simulate real-world variability and challenges encountered in clinical settings.\n\nThe authors highlight the limitations of traditional medical benchmarks that often simplify clinical scenarios and position MedFuzz as an advancement towards “beyond-the-benchmark” evaluations. Specifically, the paper presents experiments assessing LLMs' responses to MedFuzz perturbations and evaluates the consistency of chain-of-thought (CoT) explanations under these conditions. The study offers a new perspective on testing LLM robustness by addressing potential risks in clinical decision-making when assumptions of canonical benchmarks do not hold." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The paper defines robustness as the model’s ability to maintain performance in varied scenarios, which may lead to confusion with the concept of “generalization.” Typically, robustness refers to a model's resilience to perturbations or intentional adversarial attacks. To clarify the core aim of this study, a more explicit definition of robustness in the context of MedFuzz is recommended, particularly regarding how MedFuzz is designed to evaluate LLM robustness beyond generalization. Explaining how robustness is measured and differentiated from generalization could provide readers with a clearer understanding of the intended contribution.\n2. MedFuzz incorporates specific patient characteristics (e.g., age, gender, race, family history, background) as perturbations to assess LLM robustness; however, this approach may not accurately reflect clinical settings. Patient background information typically aids diagnostic decisions rather than introducing confusion. For instance, a patient’s age or medical history often plays a crucial role in diagnosis and would rarely be considered extraneous. Thus, further justification on why these characteristics are appropriate for simulating robustness under MedFuzz is recommended. Clarifying which patient data might clinically support decisions versus truly confuse the model would strengthen the study’s validity.\n3. The scale of text modification applied in MedFuzz risks excessive deviation from the original context, potentially impacting the robustness assessment. In section 3.1, for instance, added text can exceed 40% of the original passage, potentially leading to unintentional confusion beyond MedFuzz’s intended perturbation. A more focused perturbation approach—such as limiting changes to key sentences or reducing the proportion of added text—could provide a more accurate robustness assessment. This adjustment would align MedFuzz’s modifications closer to realistic conditions while still effectively evaluating LLM robustness.\n4. After applying MedFuzz, the Chain-of-Thought (CoT) explanations produced by the LLM were noted to omit important information, suggesting reduced fidelity. However, it is unclear whether this reduction in fidelity is due to MedFuzz’s perturbations or the LLM’s inherent limitations. It is recommended to first assess the fidelity and consistency of CoT explanations on the original benchmark without MedFuzz to identify the root cause of CoT discrepancies. Such an analysis would clarify whether the fidelity issues stem from MedFuzz or from the model itself, providing clearer insights into the reliability of the CoT explanations in real-world scenarios." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "An automatic redteaming method for testing the robustness of LLMs in medical question answering" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024medfuzz,\ntitle={MedFuzz: Exploring the Robustness of Large Language Models in Medical Question Answering},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=00ezkB2iZf},\nnote={under review}\n}" }, "abstract": { "value": "Large language models (LLM) have achieved impressive performance on medical question-answering benchmarks. However, high benchmark accuracy does not imply robust performance in real-world clinical settings. Medical question-answering benchmarks rely on assumptions consistent with quantifying LLM performance but that may not hold in the open world of the clinic. Yet LLMs learn broad knowledge that could help the LLM perform in practical conditions regardless of unrealistic assumptions in celebrated benchmarks. We seek to quantify how robust LLM medical question-answering benchmark performance is to violations of unrealistic benchmark assumptions. Specifically, we present an adversarial method that we call MedFuzz (for medical fuzzing). MedFuzz attempts to modify benchmark questions in ways aimed at confounding the LLM. We demonstrate the approach by targeting unrealistic assumptions about patient characteristics presented in the MedQA benchmark. Successful \"attacks\" modify a benchmark item in ways that would be unlikely to fool a medical expert but nonetheless \"trick\" the LLM into changing from a correct to an incorrect answer. Further, we present a non-parametric test for calculating the statistic significance of a successful attack. We show how to use calculate \"MedFuzzed\" performance on a medical QA benchmark, as well to find individual cases of statistically significant successful attacks. The methods show promise at providing insights into the ability of an LLM to operate robustly in more realistic settings." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "large language model", "adversarial machine learning", "automatic red teaming" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/9086aab30bbc4180cbbf3c113e82c12eecdff119.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "MedFuzz: Exploring the Robustness of Large Language Models in Medical Question Answering" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. For table 4, can you provide the same results, but for your model instead of only for TimeLLM? It would make it more obvious whether your model succeed on those tasks with incorrect textual information.\n2. For real world dataset, was the textual information always constant (as shown in section B.3) for each dataset? This would allow a finetuned model to fully ignore it, since it could bake said information in its weights anyway." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. It is good that zero shot examples of descriptions which have not been provided in the training set have been tested with. Without those, the narrow set of possible descriptions could have made it impossible to check whether the result quality came from the model overfitting on these descriptions or not.\n2. Training the model using generated data and computing how well the model follows the instructions is a relatively clean way to do a proof of concept of the idea, which is appropriate currently, as the field of using LLM and timeseries models together is still in its infancy." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The article describe a new model to incorporate textual information with a more traditional timeseries forecasting model. It does so by combining an embedding computed from the historical numerical data with an embedding computing from the textual information. The combined embedding is then used to generate the forecast.\n\nThe model is tested both on real-world data, where it shows competitive results, and on generated data, where it is shown to follow the instructions included in the textual information." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. There seems to be a mismatch between the described technique used to apply the modification (equation 3), and the examples shown (figure 3). According to the equation, the data in the forecast window should be a pure affine function, without any of the noise shown in figure 3.\n2. While the model is tested against other multimodal text+timeseries models, it should also be tested against pure LLM approaches: just plugging the text and the history in a prompt for GPT 4 or LLama 3, and looking at the generated output. While such an approach won't scale to long series, recent work have shown it to be surprisingly decent at forecasting under textual instructions. See: LLM Processes by Requiema 2024 for a slightly more complex approach, but there may be more appropriate references for the more direct one.\n3. Hyperparameters and training curiculum for the timeseries portion of the model are missing." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "- **How would the proposed model perform without access to textual inputs or under noisy conditions?** If textual instructions are incomplete, inconsistent, or contain noise, how would the model's performance be affected? This scenario is particularly relevant in high-stakes areas like finance, where decision-making often involves dealing with imperfect information. What measures have been taken to ensure robustness against these issues, which are common in real-world data?\n- **How does the proposed framework address interpretability in practice?** The paper claims that incorporating textual instructions enhances interpretability, but there are no concrete demonstrations of how this contributes to meaningful insights for domain experts. Could you provide explicit examples or user studies that validate this claim? Without such evidence, how can the claim of improved interpretability be substantiated?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- A novel two-stage framework for integrating temporal and textual data.\n- A data generation workflow for instruction-based forecasting, compatible with LLMs.\n- Comprehensive ablation studies and comparative evaluations demonstrating the effectiveness of TITSP." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents Text-Informed Time Series Prediction (TITSP), a multimodal framework that integrates textual context with time series data using Large Language Models (LLMs). The approach involves two stages: AutoPrompter, which aligns time series data with text embeddings, and a refinement stage that incorporates task-specific textual instructions to enhance prediction accuracy and interpretability. While TITSP proves particularly effective for context-rich forecasting tasks, by demonstrating improved performance under specific settings against some other methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- **Technical Contributions are Incremental** The proposed approach lacks significant technical innovation. Integrating LLMs with time series is an incremental step rather than a groundbreaking contribution. The use of cross-attention and VQ-VAE offers no substantial improvement beyond established techniques.\n- **Poor Structure and Clarity** The paper is poorly organized, with unclear explanations and an incoherent flow. The motivation and rationale for the proposed method are inadequately communicated, and critical components like AutoPrompter are explained in a convoluted manner, hindering comprehension.\n- **Inadequate Experiments** Experimental validation is weak, relying heavily on synthetic datasets that limit the assessment of practical applicability. Comparisons to related state-of-the-art methods are lacking, and statistical significance testing is absent, making it difficult to validate the performance claims.\n- **Superficial Related Work** The related work section lacks depth and fails to properly differentiate the contribution from prior research. Key works are missing or insufficiently discussed, weakening the justification for originality.\n- **Numerous Typos and Lack of Polish** Frequent typos (e.g. citation mistaches in line 54-55), poorly formatted figures(fig. 6), and poorly constructed tables suggest a lack of careful proofreading, which detracts from the overall quality and credibility of the paper.\n- **Insufficient Practical Insights** The claimed interpretability through textual integration lacks demonstration. There are no real-world examples showing how domain experts would benefit from these insights, making the practical value of TITSP unclear." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "The paper does not raise any significant ethical concerns." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see the weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The paper presents a novel approach to time series forecasting by integrating textual instructions, which is a creative extension of existing multimodal time series models. The introduction of a two-stage framework and the focus on instruction-based forecasting address a significant gap in the field.\n2. The paper is well-written and logically organized. The figures and tables are clear and effectively support the text. The problem formulation and the description of the methodology are easy to follow." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces Text-Informed Time Series Prediction (TITSP), a novel two-stage framework that enhances time series forecasting by integrating domain-specific textual information. The paper demonstrates that TITSP significantly outperforms traditional and existing multimodal approaches, improving both predictive accuracy and interpretability." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Given the synthetic data generation process, how can the authors ensure that there is no data leakage between the text data and forecasting targets? Could the authors provide a detailed explanation of the data generation process to address this concern?\n2. How practical is the proposed approach in real-world scenarios where textual instructions may not always be available or may be ambiguous? Could the authors discuss the potential limitations and challenges in deploying TITSP in practical applications?\n3. Has the model been tested on any other multimodal time series analysis tasks beyond forecasting? If not, what are the potential challenges in applying TITSP to other tasks?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Questions:\n1. The choice of order compliance rate as an evaluation metric is intriguing. This metric appears specifically tailored to the instructions outlined in the paper, which may limit its applicability to real-world scenarios. Could you clarify the advantages this metric offers over existing metrics for evaluating forecasting performance?\n\nSuggestions:\n\n- Benchmark results against a broader selection of existing multimodal forecasting models to enhance comparative insights.\n- Include a detailed discussion of the dataset, covering aspects such as sample size, history length, and forecasting horizon.\n- If feasible, incorporate more complex textual cues in the experiments to better reflect real-world forecasting challenges." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The strengths include the relevance of the problem of text-aided forecasting and the novelty of the prompting method. The methodology section is comprehensive and well-described, and the techniques and experiments have been explained in detail and are easy to follow. The figures convey the overall idea and highlight the improvements over the no-instruction setup." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes a novel two-stage for multimodal forecasting through historical data and textual cues that are useful for LLM-based forecasters. The multimodal framework is evaluated on numerous multimodal forecasting tasks. The paper provides a setup to include expert opinions for a forecasting problem." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The primary weaknesses of the paper are as follows:\n\n1. **Incomplete Literature Coverage**: Section 2.2 does not fully address relevant multimodal forecasting models, omitting key references such as UniTime ([https://dl.acm.org/doi/10.1145/3589334.3645434](https://dl.acm.org/doi/10.1145/3589334.3645434)).\n\n2. **Limited Comparative Analysis**: The results lack sufficient comparison with other multimodal forecasting models, reducing insight into how the proposed method performs relative to similar approaches.\n\n3. **Insufficient Dataset Description**: Essential dataset details, including sample counts, history length, and forecasting horizon, are not provided. Additionally, the impact of the forecasting horizon on prediction quality remains underexplored.\n\n4. **Simplistic Experimental Instructions**: The experimental instructions are overly simplistic, failing to reflect realistic scenarios. The limited set of training instructions may also suggest that simpler alternatives for instruction embedding could have been more effective.\n\n5. **Circular Evaluation**: The evaluation datasets have been tailored from existing datasets based on the training instructions intended for evaluation, which creates a circular reasoning issue that undermines the reliability of the evaluation setup. A similar statement about the order compliance rate metric can also be made.\n\n**Minor Issues:**\n\n1. The paper inconsistently uses closing quotes (\") instead of opening quotes (``) in multiple locations, including but not limited to lines 197, 203, and 213.\n\n2. Textual citations, rather than parenthetical citations, would be more suitable for the references in lines 117 to 128, enhancing the readability and flow of the text.\n\n3. Appropriate citations are not provided for the original dataset sources." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose TITSP, a multimodal framework that integrates textual knowledge with time series data using LLMs, significantly enhancing prediction accuracy and interpretability." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024instructionfollowing,\ntitle={{INSTRUCTION}-{FOLLOWING} {LLMS} {FOR} {TIME} {SERIES} {PREDICTION}: A {TWO}-{STAGE} {MULTIMODAL} {APPROACH}},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=01wMplF8TL},\nnote={under review}\n}" }, "abstract": { "value": "We introduce Text-Informed Time Series Prediction (TITSP), an innovative multimodal framework that integrates textual knowledge with temporal dynamics using Large Language Models (LLMs). TITSP employs a two-stage process that bridges numerical data with rich contextual information for enhanced forecasting accuracy and interpretability.In the first stage, we present AutoPrompter, which captures temporal dependencies from time series data and aligns them with semantically meaningful text embeddings.In the second stage, these aligned embeddings are refined by incorporating task-specific textual instructions through LLM. We evaluate TITSP on several multimodal time series prediction tasks, demonstrating substantial improvements over state-of-the-art baselines. Quantitative results reveal significant gains in predictive performance, while qualitative analyses show that textual context enhances interpretability and actionable insights. Our findings indicate that integrating multimodal inputs not only improves prediction accuracy but also fosters more intuitive, user-centered forecasting" }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Large Language Models", "Time-series Prediction", "Multi-modal", "Instruction-following" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/448e8a13abf683caa4fdc433d298a04dcb59bbe8.pdf" }, "presentation": null, "primary_area": { "value": "learning on time series and dynamical systems" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/4e0c464af7a349b9a73543bcd65624333bc923af.zip" }, "title": { "value": "INSTRUCTION-FOLLOWING LLMS FOR TIME SERIES PREDICTION: A TWO-STAGE MULTIMODAL APPROACH" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- The authors claim their approach falls under the model extraction prevention defense category. Still, it works like a detection approach where the OOD detector is built into the model itself and thus relies heavily on the OOD data used for classification. The results shared by authors, to argue otherwise, are insufficient. I would ask the authors to include more experiments for this argument. \n- If the model is trained to early exit in the case of OOD samples, but the labels used are from the original neural network (essentially the last possible exit), what is the accuracy of the model on OOD data used for training the model? I suspect that the early exit model misclassifies OOD data with high confidence. If it were learning the original network’s output labels for OOD data, then the defense would not work for the hard-label setting as the attacker would still receive a large portion of the original network’s labels as output with some erroneous ones.\n- Regarding the exit point evaluation ablation study, I would like to know the accuracy at each exit and the exact number of ID and OOD samples passing through each exit instead of terms such as “over half,” etc." }, "rating": { "value": 1 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The proposed idea of implementing early exits as a defense against model extraction is novel and sound.\n- The method is easily adaptable to different architectures like ResNets and ViTs. \n- The use of entropy and information bottleneck theory is sound and well-suited to the goal of reducing extractable information for the attacker.\n- The experiments conducted cover various scenarios, models and datasets validating its generalizability. The performance comparisons with state-of-the-art defenses further strengthen its credibility. \n- The ablation study is thorough and captures various scenarios that highlight the effectiveness of the proposed method and its components." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents “Dynamic Neural Fortress” or DNF framework as a defense against Model Extraction Attacks. These attacks allow an adversary to create a copy of a pre-trained model accessible via black-box APIs, posing risks to proprietary models. The authors identify two main challenges in current defenses: (1) Neural Network architecture protection, a thing that is taken for granted in previously proposed attacks by using the same model architecture for victim and clone models, and (2) optimizing computational resources by avoiding allocation of equal resources to both benign and attack queries. \n\nThe authors implement an Early-Exit neural network wrapper (EENN) on top of a trained model. This wrapper facilitates random exits at earlier layers for attack queries while preserving model utility by making benign queries exit at later layers. The authors assume the usage of out-of-distribution (OOD) data by attackers in most cases, but there are some experiments conducted for in-distribution (ID) data as well. Using concepts from deep information bottleneck theory, the authors optimize mutual information between input data, latent features, and output labels for training the EENN model. \n\nThe proposed method has been evaluated via testing on various architectures and datasets, and compared against other state of the art defenses." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The paper presents a technically sound idea, but the presentation is poor and needs major revisions. I am listing the weaknesses sectionwise. \n### Related work:\n- The related work is not organized properly, and some works are not cited in their appropriate sections, although they are cited later in the paper. For example, ActiveThief by Pal et al. (2020) [1] should be present under functionality stealing. \n- When a model extraction attack is data-based, the data might be natural or synthetic. For E.g., I can generate a dataset of 10,000 images from a pretrained generative network and use that for model extraction. This would still fall under the category of data-based model extraction. Data-free model extraction means that the data used for stealing is generated based on some information received from the victim. \n- Therefore, restructuring the related work section is necessary here. \n\n### Methodology:\n- The steps followed to convert a pre-trained victim model into an EENN are not easily followed. A network is trained on the ID data first. Then exit classifiers are added on top of it. Then, an OOD generator is used to generate OOD data, which is then passed through the original network without the exit networks for inference. The steps followed after this are not written in a coherent manner. One has to go through Algorithm 1 to get a clear picture of the training process.\n- Overuse of the term specific to start two consecutive paragraphs (224-235 and 236-241) and even inside the paragraphs when the sentences contained in both paragraphs are not specific at all. \n\n### Experimentation:\n- The authors should differentiate between the DFME and DBME settings in more detail. In line 387, it is assumed that the reader will know that they are talking about the DFME setting instead of the soft-label setting. This also invites confusion regarding the budget difference between the soft and hard label settings, where the budget should be the same for valid comparison. \n- For the DFME setting, one clone model architecture should be the same as the victim model for valid comparison (Resnet-34 in this case). Previous methods, like the prediction poisoning [2] method used by authors for comparison, have conducted experiments that keep the victim architecture for the stolen model. Moreover, the proposed method is not better than MeCo for the CIFAR-10 dataset. This should be analyzed and discussed.\n- For the DBME setting, using the random strategy for sampling images is not ideal. It has been shown in the ActiveThief [1] paper that using an uncertainty-based sampling method is more effective. \n- To showcase the effectiveness of the in-distribution defense, using JBDA as the attack strategy is fairly obsolete, and the paper cited needs to be corrected. The paper that proposed the attack is [3]. The authors should use either ActiveThief or Knockoff nets attack for evaluation as they are more recent and utilize intelligent sampling-based strategies for attack. If an actual attacker has access to in-distribution data, they will try to use the best strategy possible. \n- To demonstrate the defense’s effectiveness against model architecture stealing, the authors pick the latest attack by Carlini et al. but fail to show effectiveness against previously cited work, specifically “Towards reverse-engineering black-box neural networks. In International Conference on Learning Representations, 2018.” that perform attack on imagenet models. Considering that this was one of the major claims made by the authors, they should evaluate this aspect thoroughly. \n\n\n### Grammar:\nThe paper has incoherent paragraphs, spelling mistakes, and redundant sentences. Some of them are listed below:\n- Line 225, it should be “convert” instead of “covert.”\n- In Table 1 and Table 2, the spelling of label is incorrect. \n- Appendix D, Lines 778-779, same line repeated twice. \n\nCitations:\n- [1] Pal, Soham, et al. “Activethief: Model extraction using active learning and unannotated public data.” Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 34. No. 01. 2020.\n- [2] Orekondy, Tribhuvanesh, Bernt Schiele, and Mario Fritz. “Prediction poisoning: Towards defenses against dnn model stealing attacks.” arXiv preprint arXiv:1906.10908 (2019).\n- [3] Papernot, Nicolas, et al. “Practical black-box attacks against machine learning.” Proceedings of the 2017 ACM on Asia conference on computer and communications security. 2017." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Can you provide a formal definition or description of in-distribution and out-distribution data in this paper's setting? How to distinguish the normal user data (OOD) and attack data (OOD)?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "+ Good motivation. The authors adopt multi-exit architecture to defend architecture extraction attack, which is a well motivated and interesting idea.\n+ Extensive evaluation. The authors not only evaluate the defense effectiveness but also adaptive attacks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces a new defense against model extraction attack for model architecture and model utility. The key idea is to use multi-exit neural network architecture and its random exit mechanism to protect the network's architecture while ensuring the efficiency. For benign queries, the authors trains the early-exit model to distinguish OOD data (attack queries) and in-distribution data to ensure the model utility.\nFinally, the authors show that DNF outperforms previous defenses and evaluate the adaptive attacks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The assumption of attack data are OOD data, although widely adopted in prior work, should be more carefully justified. Meanwhile, as the model's training data are unknown to the user, benign queries may also be OOD data. DNF might decrease the model utility in this case.\n- The main part of paper (Section 4) is somehow hard to follow. I would suggest the author to simplify the notations or subscripts. Moreover, I also suggest the authors to provide an overview figure to replace some descriptions.\n- Although the authors investigate the adaptive attacks, the adversary can still design more powerful attack by exploiting the multi-exit model. Please discuss more about the potential vulnerability of multi-exit architecture and compare with prior attacks on multi-exit networks.\n\n[1] Auditing Membership Leakages of Multi-Exit Networks. ACM CCS 2022.\n\n[2] Model Stealing Attack against Multi-Exit Networks. arXiv:2305.13584.\n\n[3] Mind your heart: Stealthy backdoor attack on dynamic deep neural network in edge computing. IEEE INFOCOM 2023.\n\n[4] Aegis: Mitigating Targeted Bit-flip Attacks against Deep Neural Networks. Usenix Security 2023.\n\n[5] Prediction Privacy in Distributed Multi-Exit Neural Networks: Vulnerabilities and Solutions. ACM CCS 2023." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "* Can the proposed defense be easily extended to other tasks and domains, such as object detection and NLP applications?\n\n* Does the number of exit points impact the performance of the proposed defense?\n\n* According to the design, earlier blocks are intended to reduce the model's predictive capability. However, it is notable that the ID dataset maintains high accuracy even after exiting at Exit2. This raises questions about the effectiveness of the defense mechanism. Moreover, the OOD dataset still retains 35% of its data after passing through the last two blocks. What is the observed defense effect in this case?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The first defense framework simultaneously offers three key protective benefits: protecting the functionality, and model architecture, while improving the efficiency of the inference.\n\n* An innovative design of the loss function is achieved by incorporating the Information Bottleneck (IB) theory.\n\n* The experimental design is well-structured and covers various scenarios, effectively validating the method's effectiveness." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The dynamic neural fortress (DNF) defense method introduced in this paper employs a dynamic early exit neural network to defend model extraction attacks. This approach effectively provides simultaneous protection for model functionality, network architecture, and enhanced defense efficiency against these threats. Extensive experiments demonstrate that the proposed defense method outperforms SOTA model extraction defenses in terms of both effectiveness and efficiency." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* The claims regarding the protection of model architecture are overstated. Early Exit (EE) mechanisms indeed prevent attackers from executing the entire pipeline of DNN, therefore protecting the entire model architecture information from being leaked. However, the authors fail to provide how attackers might exploit this vulnerability to steal the model architecture when executing the entire network. Furthermore, EE mechanisms typically occur in the last few layers of DNNs; therefore, while the proposed approach may protect certain layers, it only works those that are unexecuted, leaving the majority of the neural network vulnerable (if there are effective attacks that can steal the model architecture). The authors should consider discussing these limitations in a dedicated section titled \"Limitations.\"\n\n* The definitions of out-of-distribution (OOD) and in-distribution (ID) data lack clarity. It is unclear why the authors consider OOD data to be \"illegal\" while ID data is deemed \"legal,\" and the rationale behind the corresponding loss term needs further explanation. Additionally, the authors aim to minimize the mutual information between $X_{id}$ and $Z_{id}$ in Eq. (3). However, this approach could potentially compromise the overall performance of deep neural networks (DNNs). The authors should provide additional clarification on why a reduced mutual information between $X_{id}$ and $Z_{id}$ does not impact the prediction accuracy.\n\n* Table 12 indicates that queries drawn from ID dataset exit at Exit2 over 90%, while the OOD queries only exit at about 75% at the same stage. This discrepancy seems inconsistent with the motivation behind two loss terms in Eq. (3) and Eq. (4). The authors should explain this discrepancy and discuss how it impacts the effectiveness of the proposed defense mechanism. I would like to suggest the authors provide a more detailed analysis of the exit patterns for ID vs OOD data.\n\n* The explanation for choosing a specific mutual information optimization method to achieve the defense objectives lacks a deeper theoretical explanation and intuitive justification, making it challenging to fully follow the principles behind the proposed method.\n\n* The experiments conducted to protect the model architecture appear limited, which does not sufficiently demonstrate the contribution related to model architecture protection mentioned in the paper. Consider adding additional experiments and evaluation metrics specifically designed to assess the robustness of the model architecture against potential theft. \n\n* It would be advantageous to include experiments that investigate the correlation between accuracy and exit points, providing a clearer visualization of the early exit mechanism's impact. I would like to suggest a graph showing accuracy vs. exit points for both ID and OOD data or report a statistical analysis of this relationship.\n\n* It seems that all datasets utilized are classification datasets, which makes it difficult to validate the effectiveness of the proposed method in other tasks and domains.\n\n* The notations in this article have been used repetitively, e.g., $r$." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. Concepts related to entropy and IB regularization are presented with some mathematical rigor and learning objectives for both ID and OOD data are presented with entropy and IB regularization contratints; However some additional insights into potential limitations are necessary – How would the strategy perform under adaptive attacks with a much varied and increasingly sophisticated OOD spectrum? And how it would impact models that aim for domain generalizability and to incorporate that OOD spectrum into their model’s capabilities?\n2. How does this defensive method translate to multi-modal architectures like VLMs. Or multi-pipeline methods where each branch operates on different modalities? Or ML methods where different models are trained for different modalities and their outputs are combined (via some aggregation)?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper presents an interesting defenseive method to counter model extraction attacks. The paper’s novelty lies in the core idea of using a dynamic exit strategy based on the input query. While early exit strategies have been explored in the context of neural networks, their application to defensive methods is novel.\n2. The paper is well written, and the core idea is simple to understand. The language is lucid but see weakness 2, 3.\n3. The paper is well organized with a clear progression between sections. Figure 1 greatly aids clarity in trying to understand the pipeline, however see weakness 2.\n4. Experimental evaluation is robust and does seem to support the author’s claims that DNF achieve substantial reduction is successful model cloning.\n5. This paper addresses a growing concern in the space of AI/ML model deployment – protecting against model cloning and privacy and intellectual rights protection. This work does have the potential to help drive forward work in defense space for these attack types." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Model extraction is a type of attack where an attacker tries to replicate a victim model to either:\n1. Estimate the model’s parameters to emulate the model’s performance\n2. Copy the model’s architecture, to recreate the model as-is.\n3. Get protected knowledge of the training data of the victim model, to better understand the data distribution it was trained on, so that other type of adversarial attacks can be done.\n\nExisting defense strategies are costly – they do not differentiate between benign and malicious queries from an attacker and this form of defense allocates the same computational power to both. This paper provides a novel way to tackle model extraction attacks – Dynamic Neural Fortresses. \n\nThey propose an early-exit strategy wherein the victim model has built-in early exits routes that the model can take and provide outputs that are OOD from it’s expected input-output combination. If an input query matches an early-exits threshold, the model inference exits with the output at that stage." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Despite strength 5, this method can be adapted widely only after these weaknesses are addressed and questions explored.\n2. Should make better use to visual elements – probably, atleast in the appendix, add an example of what an attack query would look like, why the victim system would classify the query as attack and what the victim model’s behaviour would be on it, how early would it exit?\n3. Math is useful and helps to aid the reader’s understanding but at times also hampers readability. Especially in textual sections it breaks the flow of readers. Something that may help is condense the math and limit them to equations that can be repeatedly referenced or have a table of symbol notations that readers can refer to.\n4. Some sections could use with clearer explanations - OOD Data Learning Objective, underlying theory for Entropy and IB regularization. Maybe providing examples around mutual information or ER could help.\n5. The paper does provide some explanation about Entropy and IB regularization but could expand a little more on how mutual information reduction leads to lower predictability and can be leveraged for distinguishing between benign and malignant queries.\n6. Maybe a comparison with other information-theory based approaches such as standard adversarial training would help drive home the imminent advantages on DNF. Another set of comparisons that could strengthen the paper’s results are against other dynamic architectures (example ‘BranchyNet’).\n7. The paper uses ER to determine optimal exits from the model’s inference. However the choice of thresholds is only briefly discussed. Maybe an ablation study of various hyperparameters, exit thresholds and entropy weights could help explain the choice a certain threshold or explain the assumptions that the authors may have made." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Could you please estimate the impact of early exiting for IID samples? For instance, you might compute the misalignment in model outputs for IID samples when they exit early with respect to being forwarded into the entire network.\n- Could you please evaluate the defense against a worst-case attacker, enhancing the already implemented adaptive attacks with (partial) knowledge of the training data distribution?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The paper presents a clearly novel idea to address a very relevant issue. Indeed, to the best of my knowledge, this is the first application of a multi-exit neural network to defend against model extraction attacks.\n- The proposed network architecture can also reduce the inference time during deployment.\n- The approach is very intuitive and well-justified.\n- The reported results are promising." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In this paper, a defense against model stealing attacks (targeting either the model architecture or its functionality) based on a multi-exit neural network is proposed. The main idea is to output accurate prediction scores for ID data from the later network exits, as well as uninformative scores for OOD data from the earlier exits. To do so, for each network exit, a thresholded classifier is trained on the respective intermediate layer representation with a specifically designed loss, which maximizes the aforementioned objective using concepts from information theory. During the deployment, an exit is chosen for a sample when the maximum score of an exit classifier exceeds the respective threshold." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- 90% of IID samples exit in the first 3 exits. Although this can be viewed as a benefit (it reduces the inference time), on the other side, the defense mechanism will produce less informative outputs for those samples. The impacts of these effects should be clearly understood.\n- I appreciate the fact that the authors consider different types of attacks and try to implement adaptive ones. However, a best practice when dealing with security is to simulate a worst-case scenario against the strongest attack. This helps understand the limitations of the defense and estimate lower bounds of robustness in these settings - even if, in practice, they are unlikely to occur. In this case, the adaptive attacks should be implemented using model extraction techniques that rely on some knowledge about the training data distribution. This assumption is not too unrealistic, as it might happen that the attacker (who knows the domain on which the model is applied) is able to gather in-distribution data from public domains - for instance, if the model is a malware detector, it should be very easy to collect samples and also very likely to have some overlap between them and the training data used by the victim. In other cases, the attacker might possess a subset of or all the training data, and she could easily train its own model, but she is rather interested in reproducing the exact model functionality and reproducing its decision boundaries to build a surrogate model and use it for other attacks (like evasion ones, aka adversarial examples)." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024dynamic,\ntitle={Dynamic Neural Fortresses: An Adaptive Shield for Model Extraction Defense},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=029hDSVoXK},\nnote={under review}\n}" }, "abstract": { "value": "Model extraction aims to acquire a pre-trained black-box model concealed behind a black-box API. \nExisting defense strategies against model extraction primarily concentrate on preventing the unauthorized extraction of API functionality. However, two significant challenges still need to be solved: (i) Neural network architecture of the API constitutes a form of intellectual property that also requires protection; (ii) The current practice of allocating the same network architecture to both attack and benign queries results in substantial resource wastage. To address these challenges, we propose a novel \\textit{Dynamic Neural Fortresses} (DNF) defense method, employing a dynamic Early-Exit neural network, deviating from the conventional fixed architecture. Firstly, we facilitate the random exit of attack queries from the network at earlier layers. This strategic exit point selection significantly reduces the computational cost for attack queries. Furthermore, the random exit of attack queries from earlier layers introduces increased uncertainty for attackers attempting to discern the exact architecture, thereby enhancing architectural protection. On the contrary, we aim to facilitate benign queries to exit at later layers, preserving model utility, as these layers typically yield meaningful information. \nExtensive experiments on defending against various model extraction scenarios and datasets demonstrate the effectiveness of DNF, achieving a notable 2$\\times$ improvement in efficiency and an impressive reduction of up to 12\\% in clone model accuracy compared to SOTA defense methods. Additionally, DNF provides strong protection against neural architecture theft, effectively safeguarding network architecture from being stolen." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Model Extraction Defense" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/bf422ad6c14f7dc2ca3d5a9bb6f184542a4a40f2.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Dynamic Neural Fortresses: An Adaptive Shield for Model Extraction Defense" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Although I could follow the gist of the idea, some of the notation is not precisely defined. $p_{\\mathbb{P} \\cup x*}$. It might be clearer to skip Eq.s 3/4 and jump to Eq 5.\n1. Do you have any ideas for how to generalize this to forms of data that are not amenable to KDEs (even after applying PCA)?\n1. Section 5.3 is not clear to me. What exactly is the experiment here, and what is it supposed to demonstrate?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The idea of performing MIA on a generative model by using likelihood ratio of generated data between models with and without the targeted example is very natural and efficient. I'm not surprised that it is very effective, as demonstrated in the experiments. The paper is mostly well-written and well-motivated, and to my knowledge original." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper describes a membership inference attack on generative models. It requires a set of examples generated by the model, S, and a set of reference examples, R, presumably from the same distribution as the data the model was trained on. Then to guess whether some new point x* was part of the training data, it estimates the likelihood ratio of S between a model trained on R vs. a model trained on $R \\cup \\{x*\\}$ using two kernel density estimators. It then thresholds on the likelihood ratio. Experimental results demonstrate impressive improvements compared to baseline models, particularly when evaluated with the critical \"true positive rate at low false positive rate\" metric." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I'm afraid the specific approach of using kernel density estimators will limit the method's applicability to low-dimensional tabular datasets. I would love to see this idea generalized to higher-dimensional data, probably using something that will scale better than KDEs." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1.The manuscript lacks a clear explanation of the practical utility of applying MIA to synthetic data. It remains unclear why synthetic data was chosen as the focus, rather than real-world or other benchmark datasets. The authors are encouraged to provide references in the Related Work section to strengthen the justification for studying synthetic data specifically. Expounding on the unique relevance of synthetic data to MIA would better demonstrate the necessity and contributions of this study.\n2.Several typographical errors and repeated references are present in the reference section, such as on Line 527 and Line 729. A thorough review of the references is recommended to ensure accuracy and consistency across all citations." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper introduces the Generative Likelihood Ratio Attack (Gen-LRA), a novel membership inference attack specifically aimed at detecting privacy leakage due to overfitting in generative models. Unlike prior methods, Gen-LRA employs a likelihood ratio-based hypothesis testing approach to infer membership without requiring extensive knowledge of the model structure or parameters. By leveraging density estimation techniques, the authors assess whether synthetic data generated by a model is overfitting to specific training data points, particularly in regions with outliers. The authors demonstrate that Gen-LRA significantly outperforms existing MIA methods across various generative architectures and datasets, with particular success in scenarios with low false positive rates, highlighting the nuanced privacy risks associated with generative models." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces the Generative Likelihood Ratio Attack (Gen-LRA), a novel membership inference attack specifically aimed at detecting privacy leakage due to overfitting in generative models. Unlike prior methods, Gen-LRA employs a likelihood ratio-based hypothesis testing approach to infer membership without requiring extensive knowledge of the model structure or parameters. By leveraging density estimation techniques, the authors assess whether synthetic data generated by a model is overfitting to specific training data points, particularly in regions with outliers. The authors demonstrate that Gen-LRA significantly outperforms existing MIA methods across various generative architectures and datasets, with particular success in scenarios with low false positive rates, highlighting the nuanced privacy risks associated with generative models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The effectiveness of Gen-LRA depends heavily on accurate density estimation, which can be challenging in high-dimensional data settings. The use of kernel density estimation (KDE) or principal component analysis (PCA) for dimensionality reduction may limit applicability and accuracy. This limitation is critical because the success of the Gen-LRA method hinges on reliable density estimation, which becomes less accurate in high-dimensional spaces without significant computational expense. Inaccuracies here can undermine the method's robustness, making this the most pressing limitation.\n2. Although Gen-LRA performs well at low false positive rates, its reliance on outlier detection may lead to elevated false positives in datasets with inherently high variability or complex distributions. False positives can impair the practical applicability of Gen-LRA in privacy-sensitive contexts, as overly cautious results may lead to unnecessary restrictions on data release. \n3. Gen-LRA presumes that privacy leakage primarily stems from overfitting, potentially overlooking other forms of leakage that may not manifest as local overfitting. This could lead to incomplete privacy assessments, as the Gen-LRA approach might miss privacy vulnerabilities that do not align with the overfitting model. Expanding Gen-LRA’s scope to address other leakage types could enhance its overall utility." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "The paper focuses on membership inference attacks, which could be leveraged by adversaries to launch privacy attacks." }, "flag_for_ethics_review": { "value": [ "Yes, Privacy, security and safety", "Yes, Potentially harmful insights, methodologies and applications" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "First, I would like to point out that I am not fully up-to-date on the literature regarding membership inference attacks, especially those involving tabular data. As a result, I may be unable to assess the novelty of this work and might not be familiar with the common settings examined in recent literature.\n\n1. The paper assumes the reference data is available to the attacker. This does not seem to be very realistic to me. Section 1 discussed that a common scenario for synthetic data release is that the data owner wants to release data for open research. This implies that such data is not available to the public before that (if such data is already available, then there is no motivation or value for the data owner to release an additional dataset). That means that the attacker does not have access to the reference data either. The prior work I knew often considered attacks that do not make such assumptions (e.g., https://arxiv.org/pdf/1705.07663 and https://arxiv.org/pdf/1909.03935).\n\n The paper claims that this setting is realistic in Section 2: \"We assume this in practice because this represents a plausible scenario for the owner of S as an attacker may be able to find comparable data in the real world...\" Unfortuantely, I do not fully understand this example. It would be great if the author can explain it in more detail in the rebuttal.\n\n2. Continuing on the above point, the paper needs to make it clearer what assumptions each of the baseline methods in Section 5 make. Which of them also makes the assumption that reference data is available to the attacker? This would clarify whether the claimed improvement comes from the relaxation of the assumptions or the fundamental advances of the algorithm itself.\n\n3. The paper only evaluates the proposed algorithm on tabular data. But this is not reflected in the title and abstract. By reading only the title and the abstract, the readers might be misled to think that the paper proposes and evaluates the attack on diverse data types.\n\n I think it is important to clarify that, as the proposed approach relies on kernel density estimation, which (as discussed in the paper) does not scale well with the data dimension. (The proposed approach relies on dimension-reduction techniques to tackle the issue.) Therefore, it is unclear if such a pipeline can work well on other more high-dimensional and complicated data such as images and text. \n\n4. How do you determine the kernel size and the type of the kernel in the experiments? Is the algorithm sensitive to that?\n\n5. Section 5 mentioned that \"For Gen-LRA, we found that the choice of k can have a small impact on the performance of the attack (See Appendix A.3), we therefore use the results of the best k choice for each run as the goal for an MIA is to characterize the maximal empirical privacy risk.\" I understand that choosing the best k could help \"characterize the maximal empirical privacy risk\". However, this table is mainly for comparing between different baselines. The comparison would be unfair if you chose the best hyper-parameter for your own approach while not doing that for the baseline methods.\n\n7. The discussion in Section 6.2 is nice, but it would be more self-contained if the paper could describe how DCR works in the main text.\n\n\nOther minor questions:\n\n1. Section 1: \"We demonstrate that Gen-LRA identifies a different source of privacy leakage relative to other commonly used MIAs.\" It would be better to clarify what \"the different source\" means here. I could only understand it after reading Section 5.\n\n2. Line 116 and 117: what are M and D? These notations do not seem consistent with what was used before.\n\n3. Line 127: typo on the left quotation mark\n\n4. Line 266: missing a )" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The proposed method is simple and effective.\n\n* In general, the writing of the paper is clear.\n\n* The paper has demonstrated results on many datasets and models." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes a new approach to do membership inference attacks for tabular data generative models. The approach first estimates the distributions of (1) the reference samples plus the target sample and (2) the reference samples with kernel density estimation, and then computes the density ratio of synthetic samples over these two distributions. The intuition is that, if the target sample were used in training, the density of synthetic samples on distribution (1) would be higher. Results across various datasets and models show that the proposed approach yields better AUC-ROC and TPR at low FPRs." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* The assumption that the reference data is available to the attacker is too strong.\n\n* The title and the abstract do not reflect the scope and constraint of the method sufficiently." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Can you expand the related work to also include the shadow-modeling based MIAs? \n\n- To truly understand the contribution, could you implement the shadow-modeling based MIAs [1,2,3] as well and report their results? Right now, the Gen-LRA method seems to be better than the prior work you consider, and does so with limited assumptions for the attacker and with limited computational cost. How does this change when the attacker now (i) has knowledge of the training algorithm and (ii) has the computational resources to train shadow models? Could authors implement these shadow-model MIAs and report the results alongside Gen-LRA? This would help to position the method and its results in the literature, giving a clear understanding of the impact of certain assumptions and computational cost on the MIA results. \n\n- Similarly, the work on shadow modeling MIAs also discusses disparate vulnerability of outliers [1,2,3]. Stadler et al [1] finds outliers to be more vulnerable than randomly selected records, while Meeus et al [3] proposes a method to identify more vulnerable records. Could authors have more elaborate results for the outlier discussion (e.g. show MIA results for outliers vs random points across datasets) and relate these findings to prior work? While the fact that Gen-LRA focuses on outliers is distinct from distance-based methods, these findings might not be very different than the ones in shadow-modeling based MIAs." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- Technically novel, and interesting, way to compute the membership inference inference signal from synthetic data. The method is theoretically grounded, computationally efficient and relies on limited assumptions for the attacker. \n- They show the method to outperform a range of MIAs from the literature\n- Comprehensive evaluation of the attack across 15 datasets\n- Authors include intuitive examples (eg Fig 1 and Sec 6.2) that are well explained and help the understanding of the paper." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces Gen-LRA, a novel membership inference attack (MIA) methodology for evaluating privacy risks in synthetic tabular data. The authors propose a hypothesis testing framework that computes a likelihood ratio specifically targeted at identifying any local overfitting of the target record. The method requires minimal assumptions, just access to the released synthetic dataset and a reference dataset. They find their method to outperform baselines from the literature across 15 datasets. They further find their method to be particularly successful against outliers, in contrast with other MIAs from the literature." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "(More details see questions)\n\n- My main concern comes down to a lack of related work being discussed. A range of important works have studied MIAs against synthetic tabular data using shadow modeling [1,2,3]. While I understand that these works are computationally more expensive and additionally rely on the attacker's knowledge of the training algorithm, I find these works to be very relevant to position this paper and its findings. \n- Limited secondary insights with experimental depth. For instance, to make the claim that the method works better for outliers (especially compared to other methods), section 5.3 is mostly anecdotal. \n\n[1] Stadler, T., Oprisanu, B., & Troncoso, C. (2022). Synthetic data–anonymisation groundhog day. In 31st USENIX Security Symposium (USENIX Security 22) (pp. 1451-1468).\n\n[2] Houssiau, F., Jordon, J., Cohen, S. N., Daniel, O., Elliott, A., Geddes, J., ... & Szpruch, L. TAPAS: a Toolbox for Adversarial Privacy Auditing of Synthetic Data. In NeurIPS 2022 Workshop on Synthetic Data for Empowering ML Research.\n\n[3] Meeus, M., Guepin, F., Creţu, A. M., & de Montjoye, Y. A. (2023, September). Achilles’ heels: vulnerable record identification in synthetic data publishing. In European Symposium on Research in Computer Security (pp. 380-399). Cham: Springer Nature Switzerland." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "No further questions." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The likelihood ratio that Gen-LRA estimates is novel to my knowledge, and seems to be closer to the likelihood ratio that would be theoretically optimal than what previous work has looked at. The paper is easy to understand, and the writing is generally polished.\n\nLooking at TPR @ low FPR is good practice, and too often neglected in the MIA literature. The paper could even highlight these results further: most of the AUC-ROC scores for all methods are close to random guessing, but Gen-LRA is much more accurate than random guessing at FPR = 0.001." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes a novel membership inference attack on synthetic data generators called Gen-LRA, based on estimating a likelihood ratio between the synthetic data coming from a reference distribution vs. it coming from the reference distribution with a target point included. Gen-LRA is benchmarked againt several competing attacks on a variety of datasets, where Gen-LRA generally outperforms the competition." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Using the PCA+KDE density estimator for DOMIAS is not fully fair, since the DOMIAS paper used a more sophisticated density estimator which was found to perform better than the KDE. Of course, the same estimator could also improve the results of Gen-LRA, and PCA+KDE could be computationally cheaper, but these should be checked empirically.\n\nUsing PCA may limit the applicability of outlier overfitting detection for outliers with rare categorical values. For example, consider the detection of overfitting on datapoints of French people on the Adult dataset. PCA weights the input dimensions based on how much variance they have, so the indicator for being French would have a very low weight (<1% of the data is French). As a result, the PCA outputs would be very similar between French and non-French people, and Gen-LRA would not be able to detect overfitting affecting French people. Unless I'm completely mistaken about this phenomenon, this should be mentioned as a limitation.\n\nFor a similar reason, you should check if datapoints with high DCR score have similarities. It could be that they do, but UMAP is not considering these important. This could change the interpretation of Figure 2 that DCR does not target specific outlier regions. \n\nYou should also discuss the fact that Ward et al. (2024) report a very similar finding to your Figure 2 with their MIA. As a part of this, it would be interesting to see analogues of Figure 2 for the other MIAs used as baselines.\n\nPlease include separate results from each dataset in addition to the mean results across datasets. The datasets could have significant performance differences that the aggregation hides. I'm also not sure if the standard deviations of performance across different datasets are meaningful in any way.\n\nMinor points:\n- The paper should make the differences between DOMIAS and Gen-LRA clearer, since the methods are fairly similar.\n- It not clear what $\\mathbb{P}\\cup \\{x^*\\}$ precisely is, which makes the motivation leading to Equation 4 seem a bit handwavy.\n- Contribution 1: this sentence is a bit unclear, making it seem like the null and alternative hypotheses are the same.\n- Line 172: capitalise \"equation 4\".\n- Line 266: missing parenthesis.\n- Line 346: \"scale\" is ambiguous, I would suggest \"normalise\" if that is what you are doing.\n- Several references are missing the publication forum, for example Durkan et al. (2019), Ganev and De Cristofaro (2023)." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024genlra,\ntitle={Gen-{LRA}: Towards a Principled Membership Inference Attack for Generative Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=02DCEU6vSU},\nnote={under review}\n}" }, "abstract": { "value": "Evaluating the potential privacy leakage of synthetic data is an important but unresolved problem. Most existing adversarial auditing frameworks for synthetic data rely on heuristics and unreasonable assumptions to attack the failure modes of generative models, exhibiting limited capability to describe and detect the privacy exposure of training data. In this paper, we study designing Membership Inference Attacks (MIAs) that specifically exploit the observation that generative models tend to memorize certain data points in their training sets, leading to significant local overfitting. Here, we propose Generative Likelihood Ratio Attack (Gen-LRA), a novel, computationally efficient shadow-box MIA that, with no assumption of model knowledge or access, attacks the generated synthetic dataset by conducting a hypothesis test that it is locally overfit to potential training data. Assessed over a comprehensive benchmark spanning diverse datasets, model architectures, and attack parameters, we find that Gen-LRA consistently dominates other MIAs for generative models across multiple performance metrics. These results underscore Gen-LRA's effectiveness as an interpretable and robust privacy auditing tool, highlighting the significant privacy risks posed by generative model overfitting in real-world applications" }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Privacy", "Membership Inference Attacks", "Generative Models" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/bcad18f87958725e9b50970906e168913dcdf521.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/50c96fb68049a4bec3f129b7c7f85b812793218e.pdf" }, "title": { "value": "Gen-LRA: Towards a Principled Membership Inference Attack for Generative Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see the weaknesses." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The work show the emergence of equivariant in ensemble models\n- The work generalizes previous works where the proof relied on NTKs\n- Experiments with large ensemble of models show the emergence of equivariance" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper shows that an ensemble of models when trained with data augmentation leads to emergence of equivariance properties naturally. The results generalize over past known results based on NTKs. The theory assumes some basic assumptions on the architecture and shows that, when the initialization of the weights in an architecture has some symmetry, then, the expected architecture of the ensemble is equivariant. Experimental results with various ensembles validates the results for the C4 group of symmetries." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I have several concerns over the usefulness of the theory and the experimental results.\n\nUsefulness of theory:\n- What is the use of the theory in model design or practical use cases? Since equivariant models seems to give perfect equivariance and data augmentation techniques give approximate equivariance. So, I am wondering what is the use of ensemble technique for symmetries, especially, given that we need over 1000 models to get good equivariant results.\n- What are the advantages of the proposed technique compared to existing symmetrization and canonicalization methods [1-4] that can convert non-equivariant models into equivariant ones using techniques somewhat similar to ensemble methods but with additional transformations that looks similar to augmentation.\n\nExperimental Results:\n- Although the experimental does show that the architecture with symmetric support does give invariant output, but even the asymmetric architecture seems to be giving invariant output, questioning the usefulness of the theory. It is also discussed in the paper about the symmetric states being attractors potentially, but, it still makes the current theory not very useful.\n- Experiments are only shown for C4 symmetries\n\n[1] Basu, Sourya, et al. \"Equi-tuning: Group equivariant fine-tuning of pretrained models.\" Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 37. No. 6. 2023.\n\n[2] Mondal, Arnab Kumar, et al. \"Equivariant adaptation of large pretrained models.\" Advances in Neural Information Processing Systems 36 (2023): 50293-50309.\n\n[3] Basu, Sourya, et al. \"Efficient equivariant transfer learning from pretrained models.\" Advances in Neural Information Processing Systems 36 (2024).\n\n[4] Kaba, Sékou-Oumar, et al. \"Equivariance with learned canonicalization functions.\" International Conference on Machine Learning. PMLR, 2023." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "The results in Table 1 aren't that clear to me. In the asymmetric case where you have a symmetric initialization, shouldn't you get results that are similar to the symmetric case? Yet there is a large gap" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- It generalizes the results in Gerken & Kessel \n- The topic of invariance/equivariance is important so these results would be of interest to people in that community" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper expands the results of Gerken & Kessel that show that data augmentation produces equivariant ensembles of models using NTK, by looking at finite network sizes. They then show empirically that their theoretical results indeed hold in practice (up to sampling errors)." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "My main issue is with the writing: \n- The results presented in the main text are quite trivial, that if you start with an invariant distribution and use an invariant flow you end up with an invariant distribution. The more interesting results are in the appendix (appendix B and C)\n- You writing $\\mathcal{L} = A_\\mathcal{L} + T\\mathcal{L}$ with $T\\mathcal{L}$ the tangent space is very confusing, as tangent space is defined for a manifold and we are talking about a linear space. It needlessly complicates things as there is no need to involve differential geometry when we are working on linear spaces." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Why does the OSP not increase at initialization when ensemble size increases?\n1. From the figures, it seems like the results could improve with more epochs (also for baselines). Could you please provide results with a larger number of epochs?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper is well-structured and easy to follow.\n1. The paper extends previous results to more reasonable and applicable settings. This is a significant extension." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents a theoretical analysis showing that data augmentation can lead to equivariance in deep ensembles. The paper's main result is that under several assumptions (e.g. on initialization, architecture, etc.), deep ensembles trained with data augmentation are equivariant in mean, even when individual models are generally not. A similar result was previously presented, but the paper extends these previous results, which were primarily focused on infinitely wide NNs trained with gradient descent under full augmentation, to ensembles of finite-width trained with SGD and random augmentation.\nThe paper is mainly theoretical and validates the theoretical results through limited and small-scale empirical experiments." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I like the paper and believe it has a sufficient contribution and interesting results. However, there are several limitations stated below:\n\n1. While the assumptions for the theoretical analysis are more applicable compared to previous works, they still hold only for infinite-size ensembles. Any analysis (including empirical) on the error bounds for finite ensembles would be beneficial.\n1. While the results are important, the novelty is somewhat moderate in the sense that the emergent equivariance property of ensembles was previously proposed and the fact that the theoretical analysis heavily relies on previous works [1].\n1. From the empirical evidence, it is unclear if some of the assumptions (like symmetric initialization) are indeed necessary. The authors discuss this, but I believe it can be extended further.\n1. Empirical evaluation is limited. It would be beneficial to extend it to more settings, even by small modifications like considering cyclic groups C_k of different orders (k), different architectures, model sizes, etc.\n1. It would be beneficial to see the impact of ensemble size on the metrics in Table 1, like adding a line plot for ensemble size vs. OSP. The authors show results for different sizes, but summarizing them in one clear view would make it easier to follow.\n1. The paper could benefit from a clearer and more explicit discussion of the limitations of the results.\n1. Minor:\n - Line 37: “... a definitive question to the question…”.\n\nReference\n\n[1] Flinth & Ohlsson, Optimization Dynamics of Equivariant and Augmented Neural Networks, 2023." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We prove that ensemble models learn equivariance through data augmentation." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024ensembles,\ntitle={Ensembles provably learn equivariance through data augmentation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=02Od16GFRW},\nnote={under review}\n}" }, "abstract": { "value": "Recently, it was proved that group equivariance emerges in ensembles of neural networks as the result of full augmentation in the limit of infinitely wide neural networks (neural tangent kernel limit). In this paper, we extend this result significantly. We provide a proof that this emergence does not depend on the neural tangent kernel limit at all. We also consider stochastic settings, and furthermore general architectures. For the latter, we provide a simple sufficient condition on the relation between the architecture and the action of the group for our results to hold. We validate our findings through simple numeric experiments." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "equivariance", "invariance", "ensemble models", "data augmentation", "SGD" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/d6a8bd193bcc928733dcbba2b6319d8fcb54d671.pdf" }, "presentation": null, "primary_area": { "value": "unsupervised, self-supervised, semi-supervised, and supervised representation learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/b8e3c4c2c81cde74095c52c2c359a5d2af6cf52f.zip" }, "title": { "value": "Ensembles provably learn equivariance through data augmentation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Please share missing details as mentioned in the weaknesses\n- What are the number of image and video tokens going into the LLM? How many tokens are processed by the RQ-transformer and what is its size (the RQ-VAE paper has multiple different settings)?\n- It would be interesting to see if the vision tower training results hold for a general VAE setup instead of an RQ-VAE since that would make the results even more broadly applicable" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- The paper's most interesting contribution is the unified vision tower exploration to unify generation and understanding and the appropriate ways to train such an encoder\n- The approach is quite straightforward and the application of RQ-VAE allows for token efficiency while preserving more information\n- VILA-U is close to SOTA on visual understanding tasks (image and video) with comparable models\n- The model also fares well on image generation tasks and comes close to diffusion models" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "- The paper presents VILA-U, a unified model for language, image and video understanding + generation\n- The model is trained with an autoregressive next token prediction loss for all tasks\n- The paper explores vision encoder choices to ensure understanding and generation performance" }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The method chooses RQ-VAE for efficiency, but there isn't a discussion / results around this. How would the results look if the vision tower didn't use RQ-VAE? How important is the RQ-VAE?\n- The generated images are relatively low-resolution (256 or 384px), especially since the RQ-VAE allows for increased efficiency in tokens\n- The paper doesn't really discuss video implementation details. Video understanding and generation have a mismatch in FPS / durations they usually support, what does VILA-U support? There isn't a discussion around this.\n- The paper claims to support video generation, but there are no quantitative results around this. The two qualitative examples are also very simplistic in Figure 7." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please refer to the weaknesses section." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. VILA-U introduces a unified framework that handles both visual understanding and generation in a single autoregressive next-token prediction model. \n\n2. The model leverages a unified vision tower that uses contrastive learning to align discrete visual tokens with textual inputs, which enhances the model's visual perception and text-visual alignment capabilities.\n\n3. The experiments indicate the state-of-the-art performance of VILA-U in both image generation and understanding." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents VILA-U, a unified foundation model for visual understanding and generation that integrates image and language processing into a single autoregressive next-token prediction framework. Unlike traditional visual language models that rely on separate modules or diffusion models for generation, VILA-U employs a unified vision tower to discretize visual inputs, aligning them with textual tokens through contrastive learning. From the experiments, the authors show that VILA-U can achieve state-of-the-art performance in both image generation and comprehension." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Missing the clarification between VILA-U and other tokenization-based multimodal models, like AnyGPT [1] and SEED-LLaMa [2]. Those models also used visual tokenizers to discrete the images and trained with causal language modeling loss. I noticed the authors cite the SEED-LLaMa in the line 102, but the claim of “In this work, we design our framework based on the autoregressive next-token prediction method for visual generation and make our VLM learn to generate visual content effectively.” does not the main difference between VILA-U and SEED-LLaMa.\n\n2. One of the claimed contributions of this paper is about proposing the training strategy for the unified foundation vision tower. However, the training strategy seems similar to SEED [3], which also used contrastive loss between image embeddings and text embeddings. Can authors clarify the difference between the unified foundation vision tower and SEED?\n\n3. Comparisons with other tokenization-based multimodal models [1,2] and Emu2 [4] are missing.\n\n4. The limitation section, which is required, is missing.\n\n[1] Zhan, Jun, et al. \"Anygpt: Unified multimodal llm with discrete sequence modeling.\" arXiv preprint arXiv:2402.12226 (2024).\n\n[2] Ge, Yuying, et al. \"Making llama see and draw with seed tokenizer.\" arXiv preprint arXiv:2310.01218 (2023).\n\n[3] Ge, Yuying, et al. \"Planting a seed of vision in large language model.\" arXiv preprint arXiv:2307.08041 (2023).\n\n[4] Sun, Quan, et al. \"Generative multimodal models are in-context learners.\" Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "My biggest suggestion/question is related to the number 1 weakness described above. If the author could highlight the main contribution of the work that would make its positioning much easier. One positioning that was left out in the weakness section above is to position the work as the \"first\" in some regards. However, while autoregressive modeling of text + language is a burgeoning field, VILA-U is not the first model that performs autoregressive modeling of multiple modalities." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The unification of multiple modalities in the same architecture (with the same training objective) is a very important topic. The paper is a valuable contribution to this overall research program. In the current work, the choice of quantized image tokens for image representation makes the autoregressive modeling task more natural as the image modality is tokenized into discrete tokens much like language. This helps minimizes the amount of code development required for adapting existing LLM code bases to their multimodal counterparts.\n2. The paper performed fairly complete evaluations (image-text, video-text, text-image, ) and ablation studies that include model backbone and training objective." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper, VILA-U presents a unified framework of autoregressive multimodal generation and understanding. It achieves this by first training a vision encoder (discretized via RQ codebook) for text-conditioned image tokens (initialized from CLIP) and then training image+text data using autoregressive modeling. It presents a complete training recipe for creating autoregressive multimodal models, and the resulting model is benchmarked against a wide range of existing models across tasks (generation and understanding)" }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. It is not clear to me how to position the work in its novelty or effectiveness and this may be addressable with some rewriting. I see 3 potential angles\n 1. Training effectiveness by leveraging pretrained networks. The authors motivates the work by emphasizing that existing methods that attempt to unify multimodal generation and understanding either require significant architectural modifications to their uni-modal counterparts, or training from scratch. However, this comparison seems not to play a central role in the subsequent discussions. If the effectiveness of the proposed method is reflected in ease of training, then readers would expect to see comparison of training time/compute for comparable performances. \n 2. Effective token representation of image modality as discrete tokens: VILA-U differs from prior work in its adoption of RQ-VAE embedding for images. However, if this is the main innovation, the choice of RQ, its superiority over alternative methods, the important of discontinuous embedding of images (as compared to, for example, continuous embedding as in LaViT) will need to be elevated.\n 3. State-of-the-art performance: If the main contribution is instead just the shear effectiveness of the method. Then it should demonstrate this quantitative in the paper. Unfortunately, the comparison tables doesn’t seem to suggest that the VILA-U model is the state-of-the-art in most benchmarks. Perhaps it achieves Pareto frontier between understanding and generation tasks? Or outperforms other models for the same training compute/time? Either way I’m not clear what the main advantage of the current work is over others. \n2. The discussion around training recipe is very important and useful for practitioners. However, it lacks both quantitative and qualitative (with examples) comparisons of the different training recipes. With the conclusion seems to be use an aligned CLIP model for image encoder initialization, which doesn’t seem to be a novel finding. I would recommend either supporting the discussion with more evaluation (quantitive or qualitative, ideally both) or moving the discussion to the appendix.\n3. The paper suffers from unsubstantiated claims ( neither references nor experimental support). I've highlighted a few statements that are very important for the message in the paper below:\n - \"replacing continuous tokens with VQ tokens in VLMs usually results in a severe performance drop\"\n - \"A straightforward combination of contrastive and reconstruction loss cannot converge\"\n - \"both the rFID and Top-1 accuracy of the vision tower only serves as a medium indicator instead of directly linearly correlated to the final performance of our whole multi-modal framework.\"" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "All datasets used are public, no ethics review needed." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The solid experimental results of VILA-U have largely reignited my confidence in the autoregressive image-text unified modeling direction. However, why is there no comparison with other text-image unified modeling models such as \\textbf{MM-Interleaved, SEED, and DEEM} on image understanding tasks? Ignoring the contributions of pioneers is not advisable.\n\n2. The video generation experiments are insufficient. Why not compare with methods like \\textbf{OpenSora} and \\textbf{CogVideoX} on \\textbf{VBench}?\n\n3. The article is unclear in its expression; are the visual tokens features directly discretized by the visual encoder, or are they encoded by a large language model? I suspect it is the former.\n\n4. VILA-U claims to have lower computational complexity and to avoid misalignment. While I recognize the importance of addressing misalignment, the claim of lower complexity requires experimental support. Specifically, compared to unified autoregressive image-text modeling models, using separate models like fine-tuning Stable Diffusion can also construct end-to-end autoregressive image-text modeling, which is more efficient in training and performs better. Moreover, utilizing existing mature acceleration schemes offers fast speeds. VILA-U should emphasize more on data cleansing quality and misalignment.\n\n5. Lastly, and most critically, I hypothesize that the structural improvements of the model provide minimal benefits compared to previous autoregressive unified models, with the majority of improvements stemming from the engineered data cleansing. For instance, MMC4-Core contains 22.4M data while MMC4 has 375M, yet some research indicates that training with these two datasets yields similar outcomes. Large-scale datasets like MMC4 are of very low quality. However, using just 6M of data to achieve excellent results suggests that your data is meticulously filtered, yet the paper lacks any detail on the core contributions of data construction. Conducting experiments on the same data with other model structures like \\textbf{DreamLLM} is necessary to demonstrate the efficiency of \\textbf{VILA-U}. \n\nI will improve my rating score if my concerns are addressed." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The idea of VILA-U is very straightforward, and the experiments are solid. It significantly enhances the capabilities of end-to-end autoregressive multimodal models in visual-language tasks, bridging the gap between autoregressive multimodal models and the LLAVA series, while also excelling in image generation.\n\n2. The structure of the VILA-U paper is simple and easy to read, and the model implementation is very easy to follow." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Summary:\n\nVILA-U is a foundation model that unifies video, image, and language understanding and generation. Unlike traditional models that use separate components for different tasks, VILA-U simplifies this by employing a single autoregressive framework. This reduces misalignment and maintains near state-of-the-art performance in both understanding and generating visual language content. Key factors for its success include a unified vision tower that aligns visual and textual inputs, enhancing perception, and the ability to achieve high-quality image generation similar to diffusion models.\n\nContributions:\n\n1. VILA-U strives for an end-to-end autoregressive model that handles both visual and textual inputs through a unified next-token prediction approach. This approach eliminates the need for external components like diffusion models, simplifying the infrastructure.\n2. VILA-U is tested across a range of tasks, including image-language and video-language understanding, as well as image and video generation. It demonstrates notable improvements, particularly in narrowing the gap between autoregressive and continuous-token models in visual understanding, while also offering robust visual generation capabilities." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1.Regarding the issue of missing in context learning assessments, VILA-U has undergone extensive training on image-text sequences and can accept any interleaved layouts of images and text. Therefore, it should possess excellent contextual learning abilities. This work could be enhanced by conducting tests on its ICT capabilities.\n\n2.The description of the data curation process is not sufficiently clear, making it uncertain whether the data was meticulously selected or randomly chosen. If it is the former, I suspect that most of the improvements stem from high-quality data engineering rather than advancements in model architecture." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "VILA-U is a Unified foundation model that integrates Video, Image, Language understanding and generation. It employs a single autoregressive next-token prediction framework for both tasks." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024vilau,\ntitle={{VILA}-U: a Unified Foundation Model Integrating Visual Understanding and Generation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=02haSpO453},\nnote={under review}\n}" }, "abstract": { "value": "VILA-U is a Unified foundation model that integrates Video, Image, Language understanding and generation. Traditional visual language models (VLMs) use separate modules for understanding and generating visual content, which can lead to misalignment and increased complexity. In contrast, VILA-U employs a single autoregressive next-token prediction framework for both tasks, eliminating the need for additional components like diffusion models. This approach not only simplifies the model but also achieves near state-of-the-art performance in visual language understanding and generation. The success of VILA-U is attributed to two main factors: the unified vision tower that aligns discrete visual tokens with textual inputs during pretraining, which enhances visual perception, and autoregressive image generation can achieve similar quality as diffusion models with high-quality dataset. This allows VILA-U to perform comparably to more complex models using a fully token-based autoregressive framework." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Unified Visual Language Model", "Autoregressive Model" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/50e98a8144a1cddb2de5c13e4af3f3a5a157d4f3.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "* Reward margin and offline-reward evaluation is interesting by itself and could provide information of the effectiveness of the method, but I personally think is not as an important measurement as pairwise winrate. Could you elaborate on Section 6.1 why one should consider looking into it?\n\n* Please check the questions in weaknesses as well!" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The authors test of two LLM-as-a-Judge benchmarks as well as on a well-established classification benchmark, and their results are consistent.\n* The authors provide a theoretical explanation of why their method works effectively.\n* Showing all possible combinations at Figure 2 helped understanding what kind of online RLHF methods one should consider\n* The results are consistent across smaller models (0.5B) up to widely used scale models (8B)." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Compared to offline RLHF methods, online RLHF methods empirically show stronger performance, yet is computationally expensive, vulnerable to distribution shifts and lacks a unified framework. The authors ablate different online RLHF methods based on all possible combinations (namely, SAIL-PR, SAIL-PP, SAIL-DP) which could be useful for future work exploring online RLHF methods. Personally, it was surprising that SAIL-PP generally works on par or slightly better than SAIL-PR, which open up further research questions on what would be the optimal way to obtain preference dataset." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* As a practitioner, at least the presentation/writing wasn't clear enough to agree that SAIL provides a unified framework for those who might want to consider using online RLHF in future works. I would personally suggest adding a section explains about how one could use SAIL instead of iterative DPO methods, as well as a huge emphasis on how the provided code could be used.\n* There is a huge emphasis on trying to improve reward models (on RewardBench) to mitigated reward model overoptimization & train better LMs. I am curious if given a fixed budget/time limit, whether one should try to employ online RLHF methods or try to enhance reward models in general.\n* I would suggest adding an explanation of what is the limitation of online RLHF methods that the paper could not address. For example, it is still unclear on what is the best practice to \"whether to discard instances from a preference dataset that have a subtle difference on the preference strength\" or \"would it be beneficial to employ more models when gathering responses when consisting a preference dataset\"." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "There is a large amount of blank space below Section 6.1. Is there any missing content in this part of the paper?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. Introducing Bi-level Preference Optimization: The process of bi-level preference optimization is integrated into the modeling of online RLHF. By leveraging the unique correspondence between the reward function and the LLM policy, this approach innovatively transforms the process into an equivalent single-layer form that is easier to solve.\n\n2. Extensive Experiments on SAIL: Comprehensive and rich experiments were conducted to address the three significant challenges in online RLHF and to demonstrate the relevant applications of SAIL." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors identify three significant challenges in online RLHF algorithms: Challenge 1: the interdependence between models and data in implicit reward learning; Challenge 2: the computational complexity of bi-level optimization; and Challenge 3: the reliance on preference oracles. They propose SAIL to address these challenges. \n\nThe main contributions of the paper can be summarized as follows:\n\n1. **Unified LLM Alignment Mathematical Framework**: The authors have designed a principled online RLHF framework that provides concrete guidance for generating new responses, assuming the existence of a preference oracle.\n\n2. **Adaptive Direct Preference Optimization**: By introducing a DPO-style analysis, the authors present an efficient single-layer solution capable of effectively addressing distribution shifts and providing a scalable online preference optimization method.\n\n3. **Introduction of a Self-Improvement Mechanism**: This mechanism reduces the reliance on preference oracles.\n\n4. **Extensive Experimental Evaluation**: The experiments conducted demonstrate that SAIL significantly outperforms baseline methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Regarding the three variants of the SAIL method, Table 3 shows that in the Eval-Reward and MT-bench columns, the SAIL method performs worse than the baseline DPO. Please clarify whether these experimental results undermine the assertion that the SAIL method is superior to the baseline DPO." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "1. The paper demonstrates SAIL's efficiency with models up to 8B parameters. Could you share any considerations or expected challenges for scaling SAIL to significantly larger models, such as those with over 100B parameters?\n\n2. SAIL currently relies on the Bradley-Terry preference model. Have you considered experimenting with other preference models, and do you anticipate any impact on alignment performance if different utility functions are used?\n\n3. SAIL-DP seems to show some overfitting on in-distribution responses. Could you discuss any regularization techniques you considered or plans to mitigate this, particularly to enhance generalization to out-of-distribution data?\n\n4. Given the dependence on an initial offline dataset, how does SAIL perform in situations with minimal or noisy initial data? Are there strategies within the current framework to mitigate issues arising from a limited initial dataset?\n\n5. Could you provide more detail on the computational costs of SAIL, particularly in comparison with other RLHF approaches? How does the single-level optimization approach compare in terms of resource requirements, and what practical considerations should be kept in mind when implementing it?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "1. **Innovative Formulation**: The paper provides a novel formulation of online RLHF through bilevel optimization, enhancing computational efficiency by reducing this problem to a single-level optimization, which is a significant advancement for practical LLM training.\n2. **Effective Self-improvement Mechanism**: SAIL effectively addresses challenges related to reliance on preference oracles, making online alignment more feasible by leveraging the model's self-generated responses for iterative improvement.\n3. **Comprehensive Evaluation**: The paper includes extensive experiments that demonstrate substantial improvements in evaluation reward, win rate, and efficiency over other methods like DPO, supporting SAIL's efficacy and computational advantage.\n4. **Scalability and Adaptability**: SAIL’s approach to handling distribution shifts and reducing oracle reliance presents a promising method for more scalable RLHF applications, especially for emerging large-scale LLMs.\n5. **Detailed Experiment Design and Baselines**: The experiment section is well-structured, covering a range of metrics (reward-margin, eval-reward, win rate) and configurations (SAIL-PR, SAIL-PP, SAIL-DP), providing insights into the trade-offs and performance across different setups." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces SAIL (Self-improving Efficient Online Alignment), an approach for online reinforcement learning from human feedback (RLHF) that aims to align large language models (LLMs) with human preferences. SAIL addresses limitations in offline RLHF methods by framing online LLM alignment as a bilevel optimization problem, which it reduces to a single-level first-order optimization method to enhance computational efficiency. The approach allows for continuous model improvement by generating samples iteratively, regulating preferences, and exploring online feedback. SAIL's self-improvement mechanism enables it to reduce reliance on preference oracles, thus allowing for more scalable alignment. Empirical evaluations demonstrate significant performance improvements over standard RLHF baselines." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. **Limited Exploration of Alternative Utility Functions**: The method relies on the Bradley-Terry preference model, which may not be optimal for all RLHF applications. Future work could benefit from exploring alternative utility models that account for more nuanced preference data.\n2. **Scalability Concerns for Larger Models**: Although the paper demonstrates SAIL’s effectiveness on LLMs with up to 8B parameters, additional scaling experiments would strengthen the paper's claims about computational efficiency for significantly larger models.\n3. **Dependency on Initial Offline Dataset**: While SAIL reduces oracle dependency, it still relies on an initial offline dataset to bootstrap alignment. Further discussion on managing this dependency, especially when starting with limited labeled data, could be beneficial.\n4. **Potential Overfitting in SAIL-DP**: The paper mentions that SAIL-DP shows signs of overfitting on in-distribution responses, suggesting that the method may benefit from more refined regularization techniques to ensure robust generalization to out-of-distribution samples." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "See the weakness section." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "(1) The paper introduces a novel unified framework for online RLHF that effectively addresses the challenges of static datasets and distribution shifts.\n(2) By reducing a bilevel optimization problem to a single-level method, SAIL maintains theoretical benefits while significantly lowering computational costs, making it more practical for real-world applications.\n(3) The self-improving aspect of SAIL allows models to iteratively enhance alignment without extensive supervision, addressing the challenge of needing constant access to human preference data.\n(4) Extensive experiments validate the effectiveness of SAIL, showing substantial improvements in performance metrics compared to existing methods, thus showcasing its applicability across various datasets.\n\nI would consider rescoring if the authors can solve my concern." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper addresses the limitations of traditional reinforcement learning from human feedback (RLHF) methods for aligning large language models (LLMs) with human preferences. The authors propose a unified framework for online RLHF formulated as a bilevel optimization problem, which they simplify to a single-level method for efficiency. This approach, called SAIL, allows for continuous model improvement through online exploration and iterative refinement of preference labels, mitigating issues related to distribution shifts and reducing reliance on static preference oracles. Experimental results demonstrate significant performance gains, with SAIL outperforming state-of-the-art RLHF methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "(1) The method does not improve much in the AlpacaEval 2.0 Score. The author should give a detailed explanation. And why not use metrics like length-controlled win rate?\n(2) Authors should compare more advanced preference optimization algorithms like ORPO and SimPO. And current results are not impressive for the alignment community.\n(3) Why did the author just include MMLU as the downstream task metric? They should incorporate more tasks (eg., arc-challenge) like the similar self-improvement work SPIN (ICML24) to better illustrate their contribution.\n(4) In the alignment area, it's better to conduct experiments in the Arena-Hard benchmark since it's a common metric to evaluate the alignment ability." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We introduce SAIL, an efficient online RLHF approach that addresses distribution shift and reduces reliance on preference oracles for improved LLM alignment." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024sail,\ntitle={{SAIL}: Self-improving Efficient Online Alignment of Large Language Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=02kZwCo0C3},\nnote={under review}\n}" }, "abstract": { "value": "Reinforcement Learning from Human Feedback (RLHF) is a critical method for aligning large language models (LLMs) with human preferences. However, existing offline alignment approaches, such as DPO, IPO, and SLiC, rely heavily on static datasets of human preferences, often leading to suboptimal performance. Recent efforts in the literature have moved towards online RLHF methods, but they lack a unified framework and suffer from distribution shift issues. In this work, we formalize online LLM alignment as a bilevel optimization problem. By reducing this formulation to a more computationally efficient single-level first-order method, utilizing reward-policy equivalence, we propose SAIL (Self-improving Efficient Online Alignment).SAIL generates new samples and iteratively refines model alignment through online exploration and regulation of preference labels. This enables continuous, self-improving alignment and generalizes prior online RLHF methods as special cases. Compared to state-of-the-art RLHF methods, SAIL delivers significant performance gains, with up to 11.6\\% improvement in win rate and a 3.6-point increase in evaluation rewards, while maintaining low computational overhead." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "RLHF", "Alignment", "Online Alignment", "Self-Play" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/d5a230b3d82181e94b8d74fb961b8cc3abd38e94.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/0eb1dbe2bdc367b1d3e0552efcb28e056a4766bb.zip" }, "title": { "value": "SAIL: Self-improving Efficient Online Alignment of Large Language Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. While the paper shows good empirical performance, it's unclear what types of problems would theoretically benefit most from a port-Hamiltonian approach versus standard message passing. Could the authors provide analysis or insights about which properties of the underlying data generation process would suggest using their method?\n\n2. The authors demonstrate that adding dissipative components often improves performance, but how does the balance between conservative (Hamiltonian) and dissipative parts relate to the nature of the task? It would be valuable to see an analysis of when pure conservation might be preferable to including dissipation, and how practitioners should choose this balance for new problems.\n\n3. Given that the method is inspired by physical Hamiltonian systems, it's surprising that there are no experiments on problems with actual Hamiltonian dynamics (e.g., molecular dynamics simulations). Such experiments could help validate whether the method's conservation properties provide advantages for physically meaningful conservation laws, beyond just improving general information flow." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "Overall, the idea of using Hamilton's dynamics for GNN is attractive, though not entirely new. \nNevertheless, the paper is solid and is more general than existing approaches. \nThe improved efficiency over previous Hamiltonian GNN approaches and the practical benefits for long-range propagation make it a useful contribution to the field.\nThe experimental results are also interesting. \n\n1. Technical soundness:\n- Clear theoretical analysis of conservation properties\n- Explicit connection between Hamiltonian dynamics and message passing\n- Thorough experimental validation\n\n2. Practical value:\n- More efficient than previous Hamiltonian GNN approaches\n- Good performance on long-range tasks\n- Can incorporate different message passing schemes\n\n3. Clarity:\n- Well-structured presentation\n- Good balance of theory and empirics\n- Clear comparisons to prior work" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces port-Hamiltonian Deep Graph Networks (PH-DGN), a new framework for graph neural networks that addresses the challenge of long-range information propagation. The approach embeds message passing within a port-Hamiltonian dynamical system framework, where node states are split into position (q) and momentum (p) components coupled through skew-symmetric matrices. The Hamiltonian function incorporates graph structure through neighborhood aggregation functions, and the port-Hamiltonian extension introduces learnable dissipative components (internal dampening and external forces) that allow the network to balance conservative information preservation with task-dependent information modification. The authors provide theoretical analysis of the framework's properties, including bounds on sensitivity and gradient behavior, and demonstrate empirically that their method outperforms existing approaches on several tasks requiring long-range information propagation, including synthetic graph property prediction tasks and real-world molecular property prediction. The framework can incorporate different message passing schemes and provides improved efficiency compared to previous Hamiltonian-based graph neural networks, with experimental results showing that while the purely conservative version performs well, including the dissipative components often leads to better task performance." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Novelty is incremental:\n- Builds on existing ideas (Hamiltonian GNNs, message passing)\n- Main contribution is combining these effectively rather than fundamentally new concepts\n\n2. Technical questions:\n- The derivation of the discretization scheme could use more justification\n- Some assumptions about the structure of $W$ and $V$ matrices for explicit updates feel restrictive\n- Could better explain why port-Hamiltonian framework is more appropriate than simpler alternatives\n\n3. Empirical:\n- Some ablation studies could be stronger (e.g., analyzing impact of different dissipative terms)\n- Could better justify hyperparameter choices" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 2 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "N.A." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "**Questions**\n\n- l.161: What is the definition of anti-derivative of an activation function $\\sigma$?\n- l.223, l.228: Which norm is used in $\\| \\partial \\mathbf{b}_u(T) / \\partial \\mathbf{b}_u(T-t) \\|$? \n- l.259: *Therefore, [...]. , it holds the capability of PH-GDN to perform long-range propagation effectively*: the authors compare the upper bounds of sensitivity for the existing model (Theorem 2.3) and the proposed model (Theorem 2.4), then argue that since the latter is larger than the former, the proposed model is more effective in performing long-range propagation. However, it is difficult to claim so because there is the possibility that Theorem 2.4 only shows a looser bound than Theorem 2.3, which does not exclude the possibility that the upper bound does not properly reflect the sensitivity of the PH-GDN. It should be shown theoretically or experimentally that this upper bound adequately reflects the sensitivity of the PH-GDN.\n- l.1129: I want to clarify the setup of the Graph Transfer Task: If I understand correctly, the feature vectors of all nodes are 1-dimensional and randomly sampled from $\\mathrm{Unif}([0, 0.5))$. The target value is 0 for the source node and 1 for the target node. Assuming that this problem setup is correct, I need help understanding why this task is solvable because the model cannot distinguish source or target nodes from other nodes from feature vectors.\n\n\n**Minor Comments**\n\n- l.196: *non-dissipative (i.e., long-range)*: I think this is a slightly misleading expression. My understanding is that this paper uses non-dissipative in the sense of energy-perserving. Although non-dissipative implies long-range propagation (Theorem 2.3), they are not equivalent. In fact, this paper argues that non-dissipative PH-DGN performs better than conservative PH-DGN in predicting the LRGB dataset in some numerical experiments.\n- l.436: The use of position encoding is only mentioned in the caption of Table 2 and should be explicitly stated in the text, specifically in the setup of Section 3.4.\n- l.436: The correspondence between Table 2 and Table 5 needs to be clarified. For example, the method titled *MPNNs* in Table 2 is titled *re-evaluated* in Table 5. However, GCNII is not listed as *re-evaluated*, but as *MPNNs*. Furthermore, it is difficult to tell from the captions of Table 5 whether the authors test each listed method by themselves or is a citation of existing research. The reference should be indicated for each method in Table 5, for example, by adding a column such as *reference* to Table 5.\n- l.443: *Overall, our port-Hamiltonian framework [...] shows great benefit [...] without requiring additional strategies such as global position encoding, global attention mechanism, or rewiring techniques [...].*: I suggest clarifying how the usefulness of each strategy is shown by comparing it with existing methods. More specifically:\n - The superiority of the proposed method over global position encoding is justified by the comparison with MPNN-based models using position encoding. 　　\n - The superiority over the global attention mechanism is by comparison with the Transformer-based method.\n - The superiority of rewiring methods by comparison with Drew.\n- l.1153: The reference to Adam (Kingma & Ba, 2015) should be cited.\n- l.1343: n. layers -> the number of layers\n- l.1353: Table 6 claims that PH-DGN achieves both Hamiltonian conservation and Learnable driving forces. However, I think this is misleading because to achieve Hamiltonian conservation, the dissipative component must be removed, resulting in $\\mathrm{PH-DGN}_{\\mathrm{C}}$. It is not possible to achieve both characteristics at the same time in one architecture. Instead, two variants that achieve only one of the two are proposed." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The background knowledge of the port-Hamiltonian system is explained carefully, increasing accessibility for readers unfamiliar with this topic.\n- For conservative PH-DGN, the ability of long-range interaction is theoretically shown by giving the lower bounds of sensitivity.\n- The validity of the proposed methods for synthesis datasets is properly demonstrated." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes PH-DGN, a new GNN based on the port-Hamiltonian system to develop GNN that can solve graph learning tasks that require long-range dependencies. Two variants of PH-DGN are proposed: a conservative PH-DGN based on the Hamiltonian system and a PH-DGN based on the port-Hamiltonian system by introducing learnable dissipative terms. \nThe theoretical analyses show that the conservative PH-DGN is stable and energy-preserving as a dynamical system, and derive a lower bound for the sensitivity, implying the possibility of long-range interaction.\nNumerical experiments show that the conservative PH-DGN is energy-preserving without gradient vanishing using synthesis datasets (Section 3.1) and that the long-range interaction can be achieved in graph learning tasks that require long-range interactions (Sections 3.2, 3.3). Also, the usefulness of two variants of PH-DGN is evaluated on long-range graph benchmarks of real datasets." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- As the title indicates, the theme of this paper is the relationship between GNN and long-range propagation based on the port-Hamiltonian system. However, no theoretical guarantees on long-range propagation are given for general PH-DGNs with dissipative components.\n- The tables of experimental results could be clearer (in particular Tables 2 and 5).\n- For $\\mathrm{PH-GDN}_{\\mathrm{C}}$, which has theoretical guarantees, the prediction performance on the real dataset (long-range graph benchmark) does not outperform existing methods, and hence its practical usefulness is limited." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- How would classical GCN aggregation interact with Theorem 2.4? Can the bound be easily extended for that case?\n- In Section 3.1, you mention that the growing behavior can be controlled by regularizing the weight matrices or using normalized aggregation functions. Did you try this? How are the empirical results?\n- Have you examined the interpretability of a trained PH-DGN? In particular, do the learned non-conservative forces make sense for the associated problem?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- There is very clear motivation to this work, and it builds nicely upon other references.\n- The proposed port-Hamiltonian approach is an original and clever way to allow for non-conservative dynamics in a graph network while still maintaining long-range message passing.\n- The theoretical results for the conservative case are strong, and the motivation and interpretation for these are presented nicely.\n- The experimental setup is superb; the care taken to ensure ease of replication is applauded. Model details are presented very clearly and choices are explained well for each step of the setup.\n- A strong suite of models are compared against, with many different competitors and different approaches. The consistently favorable results provide a great deal of strength to claims of the proposed method's performance.\n- The appendices are comprehensive for both proofs and experimental setup, and made clear many of the questions I had on an initial read." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This work provides a novel methodology for the incorporation of port-Hamiltonian dynamics in graph representation learning. The central model, called a port-Hamiltonian Deep Graph Network (PH-DGN), is introduced first in a purely conservative setting using only the Hamiltonian and no non-conservative terms. Several theorems are developed to show that this conservative case leads to long-range data propagation between graph nodes, where the graph dynamics exhibit no energy loss and gradients do not vanish as the backward sensitivity matrix is bounded below. Dissipitive forces are then added to the full port-Hamiltonian model, which are two additional networks that may be trained to capture non-conservative forces. Several experiments follow, including a showcase of energy conservation and sensitivity to empirically verify theoretical work, and a graph transfer problem and graph property prediction problem to compare performance on benchmark tasks against other graph models in settings which require long-range information propagation." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The majority of the theoretical results are developed for the conservative case. This makes sense in the context, as conservative long-range message passing is stated as a goal, but I would also be quite interested to see what could be proven for the fully general port-Hamiltonian case.\n- In the explanation of Theorem 2.3, the statement that \"the final representation of each node retains its complete past\" seems somewhat strong. While I understand that the BSM result shows the influence of the entire past history on the current state, this statement as written seems to imply something stronger, and perhaps could be made more clear.\n- The dissipitive force terms are added in some experiments to great success, but the explanations of their performance are more intuitive and are not supported by hard data in the paper. There may be a great opportunity here for visualization to support the intuitive claims.\n\nThere are two very minor typos:\n- In the first paragraph of Section 2, \"node states' in graph\" has an unnecessary apostrophe.\n- In Appendix D.3, \"on a grid of n. layers\" has an unnecessary period." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024porthamiltonian,\ntitle={Port-Hamiltonian Architectural Bias for Long-Range Propagation in Deep Graph Networks},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=03EkqSCKuO},\nnote={under review}\n}" }, "abstract": { "value": "The dynamics of information diffusion within graphs is a critical open issue that heavily influences graph representation learning, especially when considering long-range propagation. This calls for principled approaches that control and regulate the degree of propagation and dissipation of information throughout the neural flow. Motivated by this, we introduce port-Hamiltonian Deep Graph Networks, a novel framework that models neural information flow in graphs by building on the laws of conservation of Hamiltonian dynamical systems. We reconcile under a single theoretical and practical framework both non-dissipative long-range propagation and non-conservative behaviors, introducing tools from mechanical systems to gauge the equilibrium between the two components. Our approach can be applied to general message-passing architectures, and it provides theoretical guarantees on information conservation in time. Empirical results prove the effectiveness of our port-Hamiltonian scheme in pushing simple graph convolutional architectures to state-of-the-art performance in long-range benchmarks." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "graph representation learning", "long-range propagation", "ordinary differential equations" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/b592c630c57b079492b46ee2305981f0576b4905.pdf" }, "presentation": null, "primary_area": { "value": "learning on graphs and other geometries & topologies" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Port-Hamiltonian Architectural Bias for Long-Range Propagation in Deep Graph Networks" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please focus on the weakness of Theorem 5.2 in the rebuttal. Specifically, please compare the value of $v\\epsilon_0$ with $\\zeta^2$ in Theorem 5.2 (or address the concern in some different ways). I am willing to raise the score if the concerns are properly addressed." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "1. This paper is generally well-written. \n2. The idea of adaptive clipping intuitively makes sense and has an excellent empirical performance in the experiments of this work.\n3. Byzantine resilience in distributed learning is an important and timely topic." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a novel strategy called Adaptive Robust Clipping (ARC) for Byzantine-resilient distributed machine learning. Empirical results show that using ARC can significantly enhance Byzantine resilience compared to the methods without clipping. Theoretical analysis of convergence is also provided to show the effect of ARC." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Although the proposed ARC strategy is generally not hard to implement and has a good empirical performance, there are major concerns about the theoretical analysis in this paper, which I specify point by point below. \n\n1. The theoretical results in section 3 show that $F\\circ ARC$ is $(f,3\\kappa)$-robust when $F$ is $(f,\\kappa)$-robust (Theorem 3.2). Although the property of $ARC$ is much better than trivial clipping (as shown in Lemma 3.1), the convergence guarantee obtained from Theorem 3.2 for $F\\circ ARC$ is worse than that for $F$ (without $ARC$). In other words, the theoretical results in section 3 show that $ARC$ has a better property than trivial clipping, but do not show that using $ARC$ can improve the convergence guarantees.\n\n2. The improvement of convergence guarantees for $ARC$ is mainly shown by Theorem 5.2. Theorem 5.2 says that when the maximum gradient norm of the initial point is not larger than $\\zeta$, using $ARC\\circ F$ can guarantee to find a point $\\hat{\\theta}$ such that the square norm of the gradient at $\\hat{\\theta}$ is not larger than $v \\epsilon_0$ in expectation. However, $v \\epsilon_0$ can be much larger than $\\zeta^2$ (which is specified in the next paragraph). Briefly speaking, the result of Theorem 5.2 can be even weaker than the conditions, which makes the theorem meaningless. \n- Since $\\xi \\leq \\min(\\frac{v}{\\Phi(G,B,\\rho)},\\xi_0)$, it is obtained that $\\xi \\leq \\frac{v}{\\Phi(G,B,\\rho)}$, and thus $v\\geq \\xi \\cdot \\Phi(G,B,\\rho)=\\xi\\cdot 640(1+\\frac{1}{B^2})^2(1+\\frac{B^2\\rho^2}{G^2}).$ Therefore, \n$v\\epsilon_0 \\geq [\\xi\\cdot 640(1+\\frac{1}{B^2})^2(1+\\frac{B^2\\rho^2}{G^2})]\\cdot[\\frac{1}{4}\\cdot\\frac{G^2(f/n)}{1-(2+B^2)(f/n)}].$ The term $\\rho^2=\\exp (2\\frac{(2+B^2)\\Delta_0}{(1-\\xi_0)G^2}L)\\zeta^2$ can be much larger than $\\zeta^2$. Thus, $v\\epsilon_0$ can be much larger than $\\zeta^2$, which will make Theorem 5.2 meaningless.\n\nOverall, the idea of ARC is interesting. The ARC method is easy to implement and has a good empirical performance. However, the theoretical analysis in the current version does not show the improvement of convergence guarantees, and can be misleading." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* How does NNM contribute to achieving the guarantees outlined in Theorem 5.2? Is it possible to attain similar results on robust aggregators without incorporating NNM?\n* Using a fixed clipping threshold can often be effective in homogeneous Byzantine settings. How does the adaptive approach perform compared to a fixed threshold in such cases?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The paper proposes an adaptive method that maintains the robustness guarantees of the aggregators it employs while improving their practical performance, especially under high heterogeneity. The authors provide valuable insights into selecting the clipping threshold, demonstrating that a fixed threshold for all workers, commonly used in practice, may be inefficient in some cases and does not meet robust criteria. They also emphasize the gap between Byzantine theory and practical applications, highlighting that existing theory may not fully capture real-world performance." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces an adaptive gradient clipping method applied to worker outputs before passing them through a robust aggregator in heterogeneous synchronous Byzantine settings. The authors address a practical issue, as they observe that while fixed gradient clipping can enhance performance in some cases, it may also impair it in others. To ensure robustness while utilizing gradient clipping, they propose an adaptive strategy that adjusts the clipping threshold according to the magnitude of worker outputs, applying clipping selectively to only a subset of them. Experimental results across various Byzantine scenarios and robust aggregators, tested on MNIST and CIFAR-10 datasets, demonstrate the effectiveness of this adaptive approach when combined with the established NNM method. The authors further support their method with theoretical guarantees." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Considering the critical role that numerical evaluation plays in supporting the paper’s claims,\n* The paper introduces an adaptive clipping approach designed to work with any robust aggregator independently of NNM. However, the numerical results primarily showcase its effectiveness only when combined with the NNM aggregator (and it is unclear if NNM was also used in Figure 6; if so, this single example may be insufficient). Since NNM has a computational complexity of $O(dn^2)$, it would be valuable to assess the performance of this approach with other robust aggregators (without integrating NNM) to explore potentially lower computational costs. For instance, the CWTM ($O(dn \\log{n})$) or the $\\epsilon$-approximation GM ($O(dn + d\\epsilon^{-2})$) might offer alternatives that may retain robustness in practice while reducing time complexity. Conducting such experiments could provide a more comprehensive evaluation and emphasize the approach’s practicality.\n* The CIFAR-10 evaluation is somewhat limited, with only one Byzantine worker out of 17. Expanding the evaluation to include a higher proportion of Byzantine workers and testing on more complex datasets could better demonstrate the method’s effectiveness in more practical scenarios." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "See section before. \n\nIn addition:\n\n-In Figure 1, C=2 for static clipping (SC) is too small. I think that is why you have a very bad performance of SC. You need to test with bigger values of C as well for SC.\n\n-Can you report the Adaptive C of ARC over steps in these plots?\n\n- Line 94, can you justify why one needs to clip this exact k number gradients? what happens if one clip less or more than this number of gradients?\n\n-The intuitive k is the number of potential malicious workers which is f?\n\n-Line 238, you require \\kappa B^2 < 1, (which means B should be small) can you give an example of loss where B is small?\n\n-Line 264, I disagree with the comment that \"ARC does not introduce a significant overhead\" especially in the case of large models. It will be good to have some experiments with runtime on the x-axis \n\n-ARC theory requires that the model is well-initialized, it will be good to assess numerically the impact of the initialization on the performance of ARC" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The main strengths are: \n\n-The authors propose Adaptive Robust Clipping (ARC), a new mechanism to enhance robustness in adversarial settings.\n\n-The authors show that ARC almost retains the theoretical robustness guarantees of existing Robust methods while enhancing their practical performance. \n\n-The authors validate ARC through several experiments." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper explores enhancing the robustness of distributed machine learning in the presence of Byzantine clients. The authors propose Adaptive Robust Clipping (ARC) that improves the robustness of Robust-DGD beyond traditional static clipping techniques. ARC dynamically adjusts clipping thresholds based on gradient magnitudes, allowing it to better counteract adversarial impacts.\nThe authors provide experiments to demonstrate that ARC improves model robustness against various attacks compared to static clipping methods as well as theory showing that ARC has a similar convergence rate as the classical ones known in the literature." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The main weaknesses are:\n\n-Increased complexity produced by ARC in practical implementation\n\n-ARC performance depends on good model initialization which may degrade the performance in the case of poor initialization.\nDid you try some experiments to assess this?\n\n-While ARC improves robustness by adaptively clipping gradients, its thresholding could risk clipping too aggressively in certain settings, potentially discarding useful gradient information." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1) Why were no experiments ran using static clipping?\n2) What is the reason for only using a network with 10 agents? Typically, networks with more agents that test for heterogeneity have a harder time than those with networks because there is a wider gap on intra and inter-class datasets.\n3) What was the reason for selecting 17 agents for the simulations in the section \"Improved robustness on CIFAR-10\"? Can you expand to more agents?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The paper introduces dynamic gradient clipping as an improvement to its static counterpart. More importantly it proves that the static approach is not robust, in accordance with the definition of robustness given in the paper. Furthermore, the authors prove the robustness of their approach and provide empirical evidence of the utility of their algorithm." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduce a variation on the static clipping technique used to overcome byzantine attacks in federated learning. Their algorithm makes the process dynamic by adapting the clipping value to the input gradients themselves; this algorithm, called ARC, or Adaptive Robust Clipping, is proved to be robust: (f, 3k)-robust. More importantly, the authors prove that static clipping breaks the standard (f,k)-robustness, which highlights the shortcomings of the empirical results demonstrated in the papers highlighted in paragraph five of the Introduction. These reasons motivate the need for a dynamic approach to gradient clipping. ARC was paired with and tested using the following techniques: coordinate-wise trimmed mean, geometric median, multi-Krum, and nearest-neighbor mixing. Various simulations were ran by the authors to show the utility of ARC, these include: simulations on varying degrees of data heterogeneity, simulations on varying f (the number of malicious agents), simulations showing the improved breakdown point. All of these simulations show how ARC can provide robustness.\n\nA current problem point is that the authors perform simulations and demonstrate against not using gradient clipping. In paragraph 5 of the Introduction, the authors clearly state that static methods are a problem that their approach, ARC, solves. Then, the authors proceed to perform simulations and do not compare their results against static clipping, but compare against no clipping. It is known, and evidenced by the cited work, that not clipping is a problem that is overcome by using some form of clipping. Therefore, results compared against not clipping yields no additional information. While the authors have shown that ARC has obvious utility that could help overcome known issues, readers cannot determine the excess utility over using static clipping. While I believe the paper holds merit, as the empirical results show, I do not believe the paper can proceed without the authors running the experiments again and showing the results with static gradient clipping. The comparison between static and dynamic clipping is the fundamental point of the paper and not having a comparison of the two makes the paper unqualified to proceed. If the authors can show those results, so readers, such as myself, can see how much improvement is gained by using a dynamic choice for clipping, then I believe the paper will contain enough merit to be accepted and to receive a higher score.\n\nAs a final, syntactic, comment, I believe the authors should move the Related Works section to an earlier spot in the paper so readers can more easily understand the historical context and how the motivation for the novel work. This swap will increase flow of understanding for the reader who will have to exert less mental effort to juggle the chronological and intellectual pieces together." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The major weakness of the paper, which is a critical one, is the complete lack of comparison of ARC versus static clipping. The authors run experiments against not using clipping; this is rendered moot by prior work and therefore is not a necessary point of comparison. The authors must go back and run the same experiments they ran with with static gradient clipping and plot that against their dynamic approach. Without doing this, it is not possible to determine the benefit their work over prior work." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024the,\ntitle={The Vital Role of Gradient Clipping in Byzantine-Resilient Distributed Learning},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=03OkC0LKDD},\nnote={under review}\n}" }, "abstract": { "value": "Byzantine-resilient distributed machine learning seeks to achieve robust learning performance in the presence of misbehaving or adversarial workers.\nWhile state-of-the-art (SOTA) robust distributed gradient descent (Robust-DGD) methods were\nproven theoretically optimal, their empirical success has often relied on pre-aggregation gradient clipping.\nHowever, the currently considered static\nclipping strategy \nexhibits mixed results: improving robustness against some attacks while being ineffective or detrimental against others.\nWe address this gap by \nproposing a principled adaptive clipping strategy, termed Adaptive Robust Clipping (ARC).\nWe show that ARC consistently enhances the empirical robustness of SOTA Robust-DGD methods, while preserving the theoretical robustness guarantees. \nOur analysis shows that ARC provably improves the asymptotic convergence guarantee of Robust-DGD in the case when the model is well-initialized.\nWe validate this theoretical insight through an exhaustive set of experiments on benchmark image classification tasks.\nWe observe that the improvement induced by ARC is more pronounced in highly heterogeneous and adversarial settings." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Byzantine resilience", "distributed machine learning" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/1edff0639d6d0d7f61b9422fb9b09ffc1b65c0b2.pdf" }, "presentation": null, "primary_area": { "value": "optimization" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/3789db05085a506e826567c48ddd96568a104516.zip" }, "title": { "value": "The Vital Role of Gradient Clipping in Byzantine-Resilient Distributed Learning" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": null, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": null, "primary_area": null, "questions": null, "rating": null, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": null, "summary": null, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": null, "withdrawal_confirmation": { "value": "I have read and agree with the venue's withdrawal policy on behalf of myself and my co-authors." } }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Could the authors provide more details on the computational trade-offs involved?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The proposed method showcases two key features—*disambiguation* and *back-verification*—that notably enhance the model's reasoning process, as confirmed by the ablation study. *Disambiguation* helps clarify problem statements at each reasoning step, reducing the likelihood of misinterpretation, while *back-verification* provides a robust mechanism to cross-check each solution against previous steps. Together, these techniques improve benchmark performance by a substantial margin." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents BEATS, a framework that enhances mathematical problem-solving in language models by introducing targeted prompting strategies that guide the model through a step-by-step approach to decompose complex problems. Furthermore, BEATS incorporates a tree search mechanism, enabling exploration of each decision step individually, which helps refine solutions iteratively. The experiments demonstrate a significant performance increase on standard benchmarks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The paper combines existing approaches, such as tree search and reflective reasoning techniques, but falls short of introducing transformative new methods. While effective, the design lacks substantial innovation in handling complex reasoning beyond prior approaches.\n \n- A significant issue lies in the increased computational cost introduced by the extra steps, including disambiguation and back-verification. Although these steps improve accuracy, their contribution to computational overhead is not quantified, making it challenging to assess the overall efficiency.\n\n- Despite mentioning computational challenges in the introduction, the paper lacks a thorough analysis of the actual cost implications. The pruning technique within tree search is minimalistic, relying on basic conditions to halt expansion rather than addressing cost at a fundamental level.\n\n- Some areas in the paper, particularly Section 2.3, contain formatting issues, such as duplicated author names." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "1. How do authors verify that the disambiguation prompt effectively resolves ambiguous problem statements? Although the ablation study indicates that this prompt improves final performance, a more detailed analysis is needed. For instance, do all problems correctly solved without the disambiguation prompt remain correct when it is applied?\n2. Which version of GPT-4 is used for evaluation? If the results are referenced from OpenAI papers or technical blogs, please provide the appropriate citations." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper presents a novel approach that combines tree search with back-verification and adaptive disambiguation to enhance the mathematical reasoning capabilities of large language models (LLMs).\n2. Ablation studies are conducted to assess the impact of key components in the proposed method, focusing on the contributions of the disambiguation and back-verification modules.\n3. The pruning in the tree search effectively reduces the problem search space, improving computational efficiency." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This work investigates both prompt-based and search-based methods to enhance the mathematical reasoning abilities of large language models. The authors improve traditional search-based methods by pruning the search tree using carefully crafted prompts. A disambiguation prompt clarifies the original problem, while two additional prompts guide reasoning steps and determine search termination. Different pruning strategies are tailored to each type of prompt. The authors also introduce a self-correction mechanism called back-verification, where LLMs validate answer candidates by concatenating them with the original problem. The method’s effectiveness is evaluated across 5 math reasoning benchmarks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The proposed approach lacks substantial novelty.\n2. The selection of baselines for comparison in search-based methods is not sufficiently justified. Zhang et al. [1] use MCTS with LLaMA3 8B (which is also used in this paper) to enhance mathematical reasoning in LLMs, achieving 96.66% accuracy on GSM8K and 58.24% on MATH, which is significantly higher than the results of this approach.\n3. Although an ablation study on the BackVerify component is included, comparisons with other verification methods are lacking. For instance, the ReST paper [2] evaluates the impact of different verifiers on performance, but similar evaluations are absent in this work.\n4. While pruning tree search is a key contribution of the paper, there is no experimental analysis on the extent to which the pruning strategy reduces search time. Additionally, comparing the total inference time with other search-based methods is essential to substantiate the advantages of the pruning approach.\n\n**References:**\n- [1] Zhang, D., Huang, X., Zhou, D., Li, Y., & Ouyang, W. (2024). *Accessing GPT-4 level Mathematical Olympiad Solutions via Monte Carlo Tree Self-refine with LLaMa-3 8B*. arXiv preprint arXiv:2406.07394.\n- [2] Zhang, D., Zhoubian, S., Hu, Z., Yue, Y., Dong, Y., & Tang, J. (2024). *ReST-MCTS: LLM Self-Training via Process Reward Guided Tree Search*. arXiv preprint arXiv:2406.03816." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- How does the diversity and quality of the training data influence the performance of BEATS, particularly in edge cases or complex problems?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- BEATS uses a unique tree search strategy with pruning and back-verification to streamline reasoning paths and verify answers, improving accuracy and efficiency.\n\n- Empirical results across multiple datasets (MATH, GSM8K, SVAMP, etc.) show notable improvement over existing methods.\n\n- The inclusion of a question disambiguation component helps clarify ambiguous problems, potentially reducing error." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents BEATS, a novel approach to improving the mathematical problem-solving abilities of large language models (LLMs). It introduces a method that combines enhanced prompting, tree search with pruning, and a back-verification technique. BEATS claims significant improvements, particularly with the Qwen2-7B model, outperforming benchmarks such as GPT-4 on the MATH dataset." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- This component, though effective, adds additional steps to the inference phase, potentially affecting efficiency in real-time applications.\n\n- The paper could benefit from a more detailed discussion of the limitations of the proposed methods and potential areas for future work, such as the impact of training data on performance.\n\n- Further discussion on how the pruning limits affect accuracy vs. computation trade-off would add valuable insight." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please also refer to the weakness section.\n1. The overall framework is based on prompt engineering, which strongly relies on the capability of LLM. Can the proposed method give such significant performance improvement when dealing with Olympiad math reasoning datasets, e.g., AIME, Olympiad?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- The challenges proposed by the authors are reasonable. These challenges can inspire future research. The proposed method combines techniques that successfully alleviate the problems.\n- The experimental results are promising. The proposed method significantly\nimproves the performance of each base model compared to the comparison\nmethods.\n- This paper is well-written and organized." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper studies the mathematical reasoning problem in aspects of prompt engineering. The authors highlight the suboptimal prompts, high costs, and ineffective verification issues, and propose a tree-search-based prompt engineering method. The experiments show that the proposed method outperforms existing methods by a margin." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The novelty of this paper is somewhat limited. For example, the back verification has already been proposed in [1]. The heuristic pruning rules, e.g., Rule (3), are also common used in math reasoning. Tree-based searching methods [2] are not new either.\n- The inference cost of each method should be reported. As the SFT and zero-shot methods usually require one inference, the proposed methods require multiple samplings, making the comparison unfair.\n- The experimental results require deeper discussion. For example, the authors mention an issue with \"ambiguous problem statements\" and introduce a prompt engineering method to address it. However, there is insufficient explanation of how having the LLM rewrite the problem itself resolves this issue, and there is no comparison between the original and rewritten versions to demonstrate the effectiveness of the LLM. Additionally, if the LLM can rewrite the problem on its own, why can't it directly solve the problem?\n\n[1] Large Language Models are Better Reasoners with Self-Verification. EMNLP\n(Findings) 2023: 2550-2575\n\n[2] Accessing GPT-4 level Mathematical Olympiad Solutions via Monte Carlo\nTree Self-refine with LLaMa-3 8B: A Technical Report." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@misc{\nsun2024beats,\ntitle={{BEATS}: Optimizing {LLM} Mathematical Capabilities with BackVerify and Adaptive Disambiguate based Efficient Tree Search},\nauthor={Linzhuang Sun and Hao Liang and Jingxuan Wei and Bihui Yu and Conghui He and Zenan Zhou and Wentao Zhang},\nyear={2024},\nurl={https://openreview.net/forum?id=03u7pbpyeN}\n}" }, "abstract": { "value": "Large Language Models (LLMs) have exhibited exceptional performance across a broad range of tasks and domains. However, they still encounter difficulties in solving mathematical problems due to the rigorous and logical nature of mathematics. Previous studies have employed techniques such as supervised fine-tuning (SFT), prompt engineering, and search-based methods to improve the mathematical problem-solving abilities of LLMs. Despite these efforts, their performance remains suboptimal and demands substantial computational resources. To address this issue, we propose a novel approach, BEATS, to enhance mathematical problem-solving abilities. Our method leverages newly designed prompts that guide the model to iteratively rewrite, advance by one step, and generate answers based on previous steps. Additionally, we introduce a new back-verification technique that uses LLMs to validate the correctness of the generated answers. Furthermore, we employ a pruning tree search to optimize search time while achieving state-of-the-art (SOTA) performance. Notably, our method improves Qwen2-7b-Instruct's score from 36.94 to 61.52 (outperforming GPT-4’s 42.5) on the MATH benchmark." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": { "value": [ "~Linzhuang_Sun1", "~Hao_Liang7", "~Jingxuan_Wei1", "~Bihui_Yu1", "~Conghui_He2", "~Zenan_Zhou1", "~Wentao_Zhang1" ] }, "authors": { "value": [ "Linzhuang Sun", "Hao Liang", "Jingxuan Wei", "Bihui Yu", "Conghui He", "Zenan Zhou", "Wentao Zhang" ] }, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Large Language Models", "Tree Search", "Back Verification" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": { "value": "sun|beats_optimizing_llm_mathematical_capabilities_with_backverify_and_adaptive_disambiguate_based_efficient_tree_search" }, "pdf": { "value": "/pdf/de578034a5813e95799dc7ad53f56e41e6921edc.pdf" }, "presentation": null, "primary_area": { "value": "applications to computer vision, audio, language, and other modalities" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "BEATS: Optimizing LLM Mathematical Capabilities with BackVerify and Adaptive Disambiguate based Efficient Tree Search" }, "venue": { "value": "ICLR 2025 Conference Withdrawn Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Withdrawn_Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "The proposed model, which simply combines existing HyperNetwork and EEGNet, lacks substantial innovation. In addition, it is difficult to confirm the advantages of the proposed model from comparative experiment results as the performance improvement of the proposed method has not been consistently demonstrated across multiple datasets. This is, the proposed model achieved improved performance on the Dreyer et al. dataset in Table 1, while its performance degraded on the BCI Competition IV IIa dataset in Table 2. Furthermore, there has been no meaningful discussion about these conflicting results." }, "rating": { "value": 1 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "This study try to address the important issue of cross-user variability and BCI illiteracy issues in the MI-EEG analysis by adopting the ability of HyperNetwork to adaptive weight generation to learn user-specific representations." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In this paper, the authors proposed a HyperEEGNet architecture by combining the conventional HyperNetwork and EEGNet to adress cross-user variability for MI-BCI systems. The authors compared the performance of the proposed HyperEEGNet with that of competing EEGNet on various publicly available MI-EEG datasets in both cross-session and cross-user conditions." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "There is no substantial innovation in proposed method combining the conventional HyperNetworks and EEGNet. \n\nThe performance improvement of the proposed method over existing EEGNet has not been consistently demonstrated across multiple datasets. This is, the proposed model achieved improved performance on the Dreyer et al. dataset, while its performance degraded on the BCI Competition IV IIa dataset. Furthermore, there has been no meaningful discussion about these conflicting results.\n\nNo comparisons were conducted with existing state-of-the-art methods that have addressed the subject variability issue." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The term \"epoch\" seems to be overloaded since they are common terms used in EEG and in machine learning but mean different things. I think it becomes unclear which meaning you use in the paper sometimes, for example here is the term \"epoch\" used in close proximity, although the former seems to mean a window of data and the latter seems to mean the number of times the HyperNet is trained on all batches: \n>• Motor imagery activity data in the form of an epoch is used to perform the binary class\nclassification with a forward pass on EEGNet with the generated weights from HyperNet.\n• Cross entropy loss is accumulated for a batch of 50 epochs, and backpropagation is performed only on HyperNet parameters. Adam optimiser with learning rate 1e-4 is used.\" \n\nIt would be better if the term \"epoch\" is better clarified when used. \n\n2. >\"For the dataset from Dreyer et al. (2023), the ”acquisition runs” from 33 participants are used for training and stratified 5-fold cross-validation is used to select the best model.\" \n\nWhat variables are changed to select the best model? Is it the model architecture? Are hyperparameters tuned at all?\n\n3. Although this passage is from the \"2.4.1 Cross-Session Condition\", it seems to imply that the train-test split is not split across sessions: \n>\"For the BCI IV IIa dataset, the data from all nine participants is divided into five folds with stratified cross-validation; each fold in the iteration is considered as a test set while the other set is split with an 80-20 ratio to choose the best-performing model on the validation set.\" \n\nThe original work for the BCI competition seems to imply that there are two sessions for each subject. Is there a reason that evaluation across sessions does not seem to be done in the current work?\n\n4. In Table 1, EEGNet without the HyperNet seems to have 4 to 6 times the standard deviation as EEGNet with the HyperNet. Is there an explanation for this, especially when compared to how the ratio of the standard deviation seems to be much closer to 1 for Tables 2 and 3? If the same test in Table 1 is run with multiple seeds and using different subjects (not just the last 9 subjects) as the test set, would we still see such high variation across multiple folds for EEGNet without the HyperNet? \n\n5. For a practical control interface during deployment, it seems unclear when the resting state would occur, which is the input used in the HyperNet. Would a separate resting state classifier have to be used? Or would some other heuristic to determine resting state be sufficient?\n\n6. What is the current state of the art in terms of performance for these two datasets for the metrics you evaluated?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper is original in that they apply a HyperNet architecture to improve the generalization capabilities of an EEG motor imagery classifier. The paper evaluates inter-subject and inter-session performance, which are both important metrics for deployment of a BCI. The authors are fairly clear in how experiments are done, although I had some questions about intersession evaluation for the BCI IV IIa dataset. The work is significant in that a new method is evaluated on EEG data and shows strong generalization performance in a dataset with 42 subjects." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper aims to show the benefits of using a HyperNet architecture to improve the generalization capabilities of EEGNet for generalization given a large dataset. The authors also use data from the resting state before a trial as a novel input for motor imagery classification." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Claims on the strength of HyperNet + EEGNet would be improved through using a more comprehensive evaluation on the Dreyer et al. dataset. Leave-N-subjects-out train-test split should be done where around a quarter of the subjects are used as test subjects each time. \n2. The HyperNet + EEGNet approach does not seem to work for the BCI IV IIa dataset (one of the two datasets tested). The authors mention that it is evidence that the method does not seem to work unless with a larger dataset. It would be better if there were another dataset that can be tested to show that the HyperNet + EEGNet approach does indeed improve classification given more than 9 subjects. Alternatively, the Dreyer et al. dataset could be evaluated while varying the number of subjects for training, e.g. 8, 16, 24, 32, etc to see if the trend of improving performance given more subjects occurs." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "What specific task-related information do you think should be included to optimize the input frequency and enhance model performance? How might this affect the model's practical deployment?\n\nWhy wasn't the optimization of resting-state EEG data representations, especially concerning brain connectivity, explored? What additional features do you believe are important for downstream tasks that current measures do not capture?\n\nWhat criteria will you use to evaluate the efficacy of HyperEEGNet in comparison to other transfer learning methods? Are there particular metrics or datasets that you consider essential for this assessment?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The model exhibits robust generalization capabilities, performing effectively in both Leave-Subject-Out scenarios and with larger datasets, demonstrating its ability to handle unseen subjects. Additionally, this approach promises reduced calibration time, which is essential for real-world BCI applications, thereby enhancing user-friendliness and practicality." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduce a new architecture called HyperEEGNet aimed at enhancing EEG-based brain-computer interface (BCI) systems. This innovation addresses issues related to lengthy calibration sessions and the limited use of resting-state EEG data in motor imagery decoding tasks. By combining HyperNetworks with the EEGNet framework, HyperEEGNet adaptively generates weights for motor imagery classification based on resting-state data. In Leave-Subject-Out scenarios using a dataset of nine participants, its performance is comparable to the baseline EEGNet. However, when applied to larger datasets with 33 participants, HyperEEGNet shows improved generalization capabilities, effectively utilizing resting-state EEG information to manage unseen subjects. The model provides strong representations in both cross-session and cross-user contexts, underscoring the value of resting-state data for tasks such as motor imagery classification. Additionally, the results indicate that HyperEEGNet has a smaller memory and storage footprint, making it well-suited for edge computing. This approach offers faster user calibration and enhances the practicality of real-world BCI applications, representing a significant advancement in the field." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The initial evaluations rely on a relatively small dataset comprising just nine participants, which may not adequately reflect the variability found in larger populations. This raises questions about the generalizability of the findings without access to more extensive and diverse datasets. Additionally, while the use of resting-state EEG data is a novel approach, the model's performance may be affected if the quality or relevance of this data varies among different users or sessions. Furthermore, incorporating HyperNetworks adds a layer of complexity to the training and tuning process, potentially necessitating greater computational resources and specialized knowledge for effective implementation. Lastly, like many deep learning models, HyperEEGNet may have limitations in interpretability, making it difficult to ascertain how specific features impact its classification decisions." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "What specific strategies were implemented to mitigate overfitting during training, especially given the observed risks at larger epoch sizes? How do you plan to validate the model's performance in cross-user scenarios beyond the training dataset?\n\nCould you elaborate on the implications of a steep learning curve and rapid convergence in just 50 epochs? What does this suggest about the model's capacity to capture complex patterns in the data?\n\nWhile the focus on two-class motor imagery classification is noted, what are the plans for extending this model to accommodate multiple classes or different downstream tasks? How do you envision addressing potential challenges in this expansion?\n\n How do you explain the performance variations among participants, particularly the discrepancy in accuracy for Participant ID 3? What insights can be gained from comparing the weights generated by the HyperNet with those from an EEGNet trained directly on activity data?\n\nWhy was the optimization of resting-state EEG data representations not explored, particularly regarding brain connectivity? What additional features do you think might be important for downstream tasks that are not captured by current measures?\n\nWhat criteria will you use to compare the efficacy of HyperEEGNet against other transfer learning approaches? Are there specific metrics or datasets that you consider critical for this comparison?\n\nWhat specific task-related information do you believe should be incorporated to optimize the input frequency and further enhance model performance? How will this impact the model's practical deployment?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "Robust Generalization: The model demonstrates strong generalization capabilities, performing well in both Leave-Subject-Out scenarios and with larger datasets, indicating its effectiveness in handling unseen subjects.\n\nReduced Calibration Time: The approach promises faster user calibration, which is crucial for real-world BCI applications, making it more user-friendly." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose a novel architecture, HyperEEGNet, to improve EEG-based brain-computer interface (BCI) systems, addressing the limitations of long calibration sessions and the underutilization of resting-state EEG data in motor imagery decoding tasks. By integrating HyperNetworks with the EEGNet architecture, HyperEEGNet adaptively generates weights for motor imagery classification based on resting-state data. In Leave-Subject-Out scenarios using a dataset with nine participants, the model performs comparably to the baseline EEGNet. However, when scaled to datasets with 33 participants, HyperEEGNet demonstrates enhanced generalization capabilities, effectively leveraging resting-state EEG information to handle unseen subjects. The model achieves robust representations in both cross-session and cross-user scenarios, highlighting the potential of resting-state data for downstream tasks like motor imagery classification. Furthermore, the findings indicate that HyperEEGNet's smaller footprint reduces memory and storage requirements, making it suitable for edge computing. This approach promises faster user calibration and improved feasibility for real-world BCI applications, advancing the field significantly." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Limited Dataset Size: The initial evaluations involve a relatively small dataset with only nine participants, which may not fully capture the variability present in broader populations. The generalizability of the findings could be questioned without larger, more diverse datasets.\n\nDependence on Resting-State Data: While leveraging resting-state EEG data is innovative, the model's effectiveness might be limited if the quality or relevance of the resting-state data varies across users or sessions.\n\nComplexity of HyperNetworks: The integration of HyperNetworks may introduce additional complexity in model training and tuning, potentially requiring more computational resources and expertise to implement effectively.\n\nInterpretability: As with many deep learning models, the interpretability of HyperEEGNet's decision-making process might be limited, making it challenging to understand how specific features influence classifications." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose an architecture that leverages resting state EEG data for generating weights via hypernetworks for downstream task like Motor Imagery" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024from,\ntitle={From Rest to Action: Adaptive Weight Generation for Motor Imagery Classification from Resting-State {EEG} Using Hypernetworks},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=04RGjODVj3},\nnote={under review}\n}" }, "abstract": { "value": "Existing EEG-based brain-computer interface (BCI) systems require long calibration sessions from the intended users to train the models, limiting their use in real-world applications. Additionally, despite containing user-specific information and features correlating with BCI performance of a user, resting-state EEG data is underutilized, especially in motor imagery decoding tasks. To address the challenge of within and across-user generalisation, we propose a novel architecture, HyperEEGNet, which integrates HyperNetworks (HNs) with the EEGNet architecture to adaptively generate weights for motor imagery classification based on resting-state data. Our approach performs similarly in a Leave-Subject-Out scenario using a dataset with 9 participants, compared to the baseline EEGNet. When the dataset size is scaled, with 33 participants' datasets, the model demonstrates its generalisation capabilities using the information from resting state EEG data, particularly when faced with unseen subjects. Our model can learn robust representations in both cross-session and cross-user scenarios, opening a novel premise to leverage the resting state data for downstream tasks like motor imagery classification. The findings also demonstrate that such models with smaller footprints reduce memory and storage requirements for edge computing. The approach opens up avenues for faster user calibration and better feasibility of edge computing, a favourable combination to push forward the efforts to bring BCIs to real-world applications." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Brain-Computer Interfaces (BCIs)", "Motor Imagery", "HyperNetworks", "Data driven learning", "Adaptive weights" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/b85b237d2a962f5758e3e5244045cb38ec1196a8.pdf" }, "presentation": null, "primary_area": { "value": "applications to neuroscience & cognitive science" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "From Rest to Action: Adaptive Weight Generation for Motor Imagery Classification from Resting-State EEG Using Hypernetworks" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Could you elaborate on the training details, such as the learning rate and optimizer? Did you properly tune them and were the results sensitive to these choices?\n- Does NanoMoE lead to higher activation memory due to larger intermediate tensors?\n- How does NanoMoE compare with low-rank in terms of performance vs wall clock time (rather than parameter count or FLOPs) ?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- NanoMoE is a novel family of structured matrices with clear theoretical advantages over low-rank matrices in terms of expressiveness, especially in achieving higher ranks for the same parameter count.\n- The paper rigorously proves the said advantages of NanoMoE\n- Experiments support the theory." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces NanoMoE, a novel family of structured matrices that achieves superior flexibility compared to low-rank matrices with minimal increase in parameters or FLOPs. The paper theoretically proves that NanoMoE can have a significantly higher rank and is strictly more flexible than low-rank matrices for similar parameter counts. Some experiments confirm the improved performance of NanoMoE layers relative to low-rank layers." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Experiments are only done on toy problems such as dense matrix approximation and a small text classification dataset. Can the authors present experiments on tasks such as image classification on CIFAR-10 / ImageNet and language modeling (e.g., using the nanoGPT codebase)? Results on these benchmarks have been used in evaluating new structured matrices in recent works [1, 2].\n- There are already many equally parameter-efficient structured matrices that have the advantage of being full-rank, such as the Kronecker product, Tensor-Train decomposition, and Monarch matrices [1]. There is no comparison with these alternatives.\n- While more expressive than the usual low-rank matrix, I believe NanoMoE will require more memory to store the activations (intermediate tensors have size $K r$ rather than just $r$). Moreover, I suspect the tensor core utilization will be lower because the block diagonal matrices involve contraction with smaller ranges, resulting in worse wall clock times despite having a minimal increase in FLOPs. The authors did not discuss these potential limitations.\n- The experiment section does not provide details about how the models were trained. For example, are the learning rates well-tuned? Prior work [2, 3] has shown that structured matrices require very different learning rates than those commonly used for dense layers, making a well-tuned learning rate important for a fair comparison.\n- The paper presents the connection to MoE as a strength since it has been shown to be more compute-efficient for pre-training LLMs. But only sparse MoE models have demonstrated improved training efficiency and is what was used in referenced models such as Mixtral and Switch Transformer. The proposed NanoMoE, however, is not a sparse MoE model and is, therefore, unlikely to lead to similar benefits. The authors carefully discuss this distinction.\n- Recent works have used structured matrices to build MoE in each linear layer, similar to what is proposed in this work. I suggest the authors to discuss these highly related works. [3, 4]\n\n[1] Dao et al. 2022. Monarch: Expressive Structured Matrices for Efficient and Accurate Training\n\n[2] Qiu et al. 2024. Compute Better Spent: Replacing Dense Layers with Structured Matrices\n\n[3] Potapczynski et al. 2024. Searching for Efficient Linear Layers over a Continuous Space of Structured Matrices\n\n[4] Oldfield et al. 2024. Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "- What are the hyper-parameters used to conduct the experiments? See weaknesses section for what it’s relevant to discuss. What’s the sensitivity of NanoMoE to hyper-parameters?\n- What is the loss function used to optimise the first experiment of section 4? What is this experiment trying to show (irrespective of matching conclusions with the AG news experiment)?\n- What is the runtime of NanoMoE compared to dense matmuls either with low-rank or not? How complicated is it to run this efficiently in modern accelerators? Is this future work?\n- Is NanoMoE more prone to overfitting in the experiments?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The method to mix low-rank factorization intermediate outputs is interesting.\n- The paper emphasizes how performance improves as a function of FLOPs, which is important to develop scalable methods, especially for pre-training." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces NanoMoE, a building block which adds an additional mixing layer to low-rank factorization layers for linear projections. The paper draws connections to the mixing matrix from the mixture of expert literature, and characterises the space of matrices it can represent. Finally the authors test the proposed method on a synthetic task on various FLOPs budgets and on the AG news classification task." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "**Conceptual Framing**\n\n- The connection to the sparsely-gated mixture of experts literature is very weak. As the method is described in section 3, the matrix M performs mixing over the partitioned input x_in, this is very different from [1] which is referenced in section 1, where specific components of the network learn to route inputs to “experts”. Nanomoe rather does a sort of sparse mixing over the embedding dimension, where no sparse routing or “expertise” is learned.\n- The claims in the paper are too overreaching.\n - Mixture of expert layers and sparse layers have already been applied to individual layers in prior work, see [1] and [2] [3] for preliminary references, a more thorough literature review should be in the paper. This work does not scale more than previous work in terms of applying them to whole components of the network, or in the experimental setting size (which is very modest in NanoMoE).\n - Section 1 says “We formally define this problem as parameter-efficient deep learning”. There is not a formal definition of this in the paper, just a formal definition of the proposed method. Also, more related to the proposed approach, is sparse deep learning, which has a rich body of literature. The paper hardly proposes a new problem that is not already known or tackled in the deep learning literature.\n - There is no discussion about hardware efficiency other than a very loose definition of FLOPs in numpy. Real-world hardware efficiency is necessary to scale up methods as implied in the title.\n- The Monarch matrices line of work [5] [6] [7] seems very relevant to this work (it is not cited), as it deals with a more efficient building blocks with block-diagonal matrices, with detailed discussion on expressiveness, efficiency, experimental work and covering a super-set of the scope of this paper (both pre training with mixed sparse+dense, and finetuning of dense models). I highly recommend the authors to review [5] as a blueprint for this work. It’s worth a discussion of the differences between both methods, both in terms of modelling and hardware efficiency for pretraining; at the very least, this seems like an important baseline to have in the experimental section.\n\n**Experiments**\n\n- I found the first experiment from the OPT-13b layer very confusing. There is no description about what loss is being optimised, which makes it very difficult to interpret the results — the losses in Figure 2) and b) seem high but without any description it is not possible to know if any of the models is learning anything useful at all. Moreover, the input is random gaussians with a rather high standard deviation, again without any explanation, this task does not seem to be representative of a real training task at all.\n- The experiments compare to Low-Rank training as a baseline. However, a more important comparison to do is with a fully dense layer, which is what actually is commonly used in pre-training (which the paper advocates for in Section 1). Also, the related work section describes a number of models that would be important baselines to compare to, low-rank is arguably a simple baseline and not SOTA.\n- For the AG News classification dataset, there’s several important experimental details missing, which are crucial to understand the empirical merit of NanoMoE:\n - What is the loss being optimised?\n - What are the details of the vectorization layer? What’s the vocabulary size? How are words out of vocabulary handled?\n - How many epochs/steps occur during training?\n - What is the optimizer and what hyper-parameters are used? (batch size, learning rate, regularisation, etc)\n - How are the weights initialised in the NanoMoE layers? More generally speaking, which hyper-parameters are different in NanoMoE vs the low-rank baseline?\n - What is the granularity of the K and r ranges?\n - What is the activation function used in the experiments?\n- [7] shows that a careful parametrization is needed for structured matrices. This is a missing detail on the hyper-parameters, but also a missing discussion for NanoMoE too.\n- Figures 4) and 5) are hard to visualise with all the data points being very transparent. There is a lot of variance per Flop Budget, which probably is due to interactions of K and r. It is important to disentangle these effects as well.\n- Plotting the low envelope seems to ignore the fact that NanoMoE is overfitting at higher FLOP counts on figure 5b (if that’s not the case, the colours are making this difficult to interpret). Is NanoMoE more prone to overfitting at higher FLOP budgets? If it is, then the method is not very promising, it could also be a lack of proper regularisation, but this is not clear given the lack of experimental details.\n- Modern NLP solves classification problems such as AG News with unsupervised pre-training + transfer-learning (BERT-style models) or few-shot learning (GPT-style models). While large-scale pre-training is very expensive, there is work to pretrain BERT-style models in as little as 1 GPU day [4] which is more suitable to academic budgets. A *single* and small-scale experiment on this unsupervised learning setup, would be more apt to compare to modern methods in NLP (this can very well be the single best combination from the AG news experiment).\n- The definition of FLOPs seem to focus on inference considerations, as I think it computes the output of numpy.einsum_path over a single einsum operation (is not clear what operations are included in the call to enisum_path, a spelled out code snippet would be useful). However, this paper focuses as per section 1 on efficient pre-training. This calls for a definition of FLOPs per training step, which includes: forward and backward FLOPs, runtime bounds such as given in [5], and practical step time on modern accelerators. A number of these can be future work, but it needs to be disclosed explicitly in order to consider the merits of the paper.\n- All in all, I consider the experimental section to be too weak to claim this in the conclusion: “our empirical results consistently validate that NanoMoE achieves superior performance”. More thorough experiments need to be done before claiming this.\n\n[1] https://openreview.net/forum?id=B1ckMDqlg\n\n[2] https://proceedings.mlr.press/v162/dao22a/dao22a.pdf\n\n[3] https://openreview.net/forum?id=-b5OSCydOMe\n\n[4] https://proceedings.mlr.press/v202/geiping23a/geiping23a.pdf\n\n[5] https://proceedings.mlr.press/v162/dao22a/dao22a.pdf\n\n[6] https://openreview.net/forum?id=cB0BImqSS9&noteId=98BZANkxc8\n\n[7] https://proceedings.mlr.press/v235/qiu24f.html" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "**Questions:**\n- How were the hyperparameters chosen? Was any analysis conducted to determine optimal values, especially for selecting the dense layer in synthetic data experiments?\n- What prevented the focus from extending to multiple FFN layers? Was this due to increased complexity, as each dense layer would require a similar setup?\n- Why has NanoMoE not been tested on more complex architectures beyond single dense layers? How do the authors envision scaling it for larger models?\n- Is there a reason NanoMoE did not incorporate sparse gating mechanisms, as seen in other MoE frameworks?\n- How does NanoMoE compare with other parameter-efficient MoE-based or low-rank models in terms of accuracy and parameter reduction? Were any qualitative comparisons made?\n- Has NanoMoE been tested in transfer learning contexts? Would it retain its efficiency and performance when adapted to new tasks?\n\n\n**Suggestions:**\nThe theoretical foundation is strong, but more experiments are needed to assess NanoMoE's performance and complexity compared to other MoE and existing approaches. I suggest:\n- Adding performance comparisons with some existing baselines.\n- Extending the experiments to more layers beyond the embedding layer, ideally including FFN and attention layers for a thorough evaluation.\n- Compared with other MoE frameworks, NanoMoE’s structure is similar and would benefit from these benchmarks." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The strengths of this paper can be summarized as follows:\n\n- Efficiency in Resource Usage: The proposed method effectively reduces the number of parameters and computational demands, making it suitable for deployment in resource-constrained environments.\n- Maintained Performance: Despite the reduction in computational resources, the model achieves results that are competitive with more resource-intensive approaches.\n- Innovative Approach: The factorization and aggregation technique offers a fresh perspective on optimizing neural network architectures." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a novel variant of the mixture of experts model aimed at reducing the number of parameters and floating-point operations (FLOPs) in neural networks. This is achieved by factorizing individual feedforward layers and their corresponding input feature maps into smaller chunks, and then aggregating their outputs. By applying this operation to dense layers, the method significantly reduces parameter count and memory requirements while maintaining competitive performance levels\n\nThe paper’s main contributions include: introducing NanoMoE, a parameter-efficient block family with three complexity levels (NanoMoE-I, II, and III); proving NanoMoE’s higher expressivity over low-rank factorization with minimal parameter increase; and validating through experiments that NanoMoE achieves better model quality than low-rank factorization at similar resource budgets." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Although the idea is interesting, the proposed method has several major weaknesses:\n\n- **Lack of Inference Speed Evaluation**: While the main objective is to reduce computational cost and memory footprint, the experiments focus primarily on parameter reduction. There is no discussion of whether the method improves inference speed, which is critical for assessing practical efficiency gains.\n\n- **Limited Experimental Scope**: The authors conduct only two experiments on a single dense layer or a simple model, making it difficult to assess the method’s feasibility for real-world deployment and its performance in more complex scenarios.\n\n- **Narrow Evaluation Metrics**: The evaluation is limited to loss reduction without considering classification accuracy, which would be valuable for classification tasks. Including transfer learning experiments would further help to gauge the method’s effectiveness across tasks.\n\n- **Absence of Baseline Comparison**: The approach of weight partitioning and non-gated mixtures of experts is not new[1]. Comparisons with existing methods that use similar techniques, such as [1] focusing on parameter reduction, would provide clearer insights into the proposed method’s relative performance and innovation.\n\n[1] Scaling Laws for Fine-Grained Mixture of Experts" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "What’s the loss function in the synthetic dataset experiments where you are sampling i.i.d gaussian random vectors of dimension 20480? The text mentions that it’s testing the FC layer from OPT-13B, which is a language model. It’s not clear to me what’s the training objective function here." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The idea is quite relevant to lots of on-going works that replace dense matrices with different structure matrices for improved performance. I think it’s a nice addition to the community.\n- The paper is well-written and easy to follow" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose to extend low-rank approximation of standard neural network weight matrices of the form W=UV into W = blockdiag(U) M blockdiag(V) where blockdiag(U) is a block diagonal reshaping of the original matrix U. M is a block matrix interpreted as expert weights for each possible combinations of subblocks from blockdiag(U) and blockdiag(V). The M matrix is parametrized in three different ways (scalar times identity, diagonal, diagonal plus outer product) with increasing expressivity proved theoretically but also increasing computational cost. The authors empirically validate that the proposed approach is better than low-rank in terms of train/test loss for 1) a synthetic data setting when controlling parameters and FLOPs, 2) AG news classification when controlling for parameters and FLOPs." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The primary problem with all of the empirical evaluations in this paper is that they are not informative about whether the proposed approach is actually a good replacement for standard MoE layers or not. The baseline is just low-rank, which is shown to be less expressive already compared to the proposed nanoMoE. It’s essential to compare to a standard MoE where you’re not using any low-rank structure but with just dense matrices. \n\nIt’s also not surprising that when controlling for parameters, nanoMoE is performing better than low-rank since the parameter overhead introduced by K and r are relatively small (the authors sweeped over small values of K). \n\nI’m willing to change my scores if the authors add the dense matrix $W\\in\\mathbb{R}^{d_2\\times d_1}$ baseline and the standard MoE with dense matrices baseline, at least in a limited setting if the compute budget is a problem during rebuttal." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024nanomoe,\ntitle={NanoMoE: Scaling Mixture of Experts to Individual Layers for Parameter-Efficient Deep Learning},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=04RLVxDvig},\nnote={under review}\n}" }, "abstract": { "value": "Large language models (LLMs) have achieved remarkable success, but their growing size leads to significant challenges in efficiency and cost. This work explores parameter-efficient deep learning, aiming to achieve comparable performance with fewer parameters and floating-point operations (FLOPs). We introduce NanoMoE, a novel family of parameter-efficient building blocks inspired by the Mixture of Experts (MoE) framework. NanoMoE offers a modular and efficient replacement for fully connected layers within traditional neural networks. We instantiate NanoMoE with three variants of increasing complexity and theoretically demonstrate its superior expressivity compared to low-rank factorization with minimal parameter increase. Empirical results validate that NanoMoE achieves superior model quality compared to low-rank factorization under the same parameter or FLOP budget, confirming its enhanced efficiency." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Mixture of Experts", "Parameter Efficiency", "Expressivity", "Low-Rank Factorization" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/4aa38d228c8912d3aa307666a911d282faf7ee53.pdf" }, "presentation": null, "primary_area": { "value": "unsupervised, self-supervised, semi-supervised, and supervised representation learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "NanoMoE: Scaling Mixture of Experts to Individual Layers for Parameter-Efficient Deep Learning" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Prompt-based continual learning methods like L2P[1], DualPrompt[2], S-Prompts[3] and HiDe-Prompt[4] employ various prompt design and selection strategies. As for the prompt design, how does this paper demonstrate the superiority of the proposed method?\n2. Is there a writing error in Equation 12? This loss aims to increase the similarity between $x^t_P$ and $x_{instruct}$; however, as $x^t_P$ and $x_{instruct}$ become more similar, it means the prompt cannot provide additional information, which would be detrimental to prompt learning.\n\n[1] Wang Z, Zhang Z, Lee C Y, et al. Learning to prompt for continual learning[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022: 139-149.\n\n[2] Wang Z, Zhang Z, Ebrahimi S, et al. Dualprompt: Complementary prompting for rehearsal-free continual learning[C]//European Conference on Computer Vision. Cham: Springer Nature Switzerland, 2022: 631-648.\n\n[3] Wang Y, Huang Z, Hong X. S-prompts learning with pre-trained transformers: An occam’s razor for domain incremental learning[J]. Advances in Neural Information Processing Systems, 2022, 35: 5682-5695.\n\n[4] Wang L, Xie J, Zhang X, et al. Hierarchical decomposition of prompt-based continual learning: Rethinking obscured sub-optimality[J]. Advances in Neural Information Processing Systems, 2024, 36." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper is well written and is the first prompt learning framework for rehearsal-free continual learning of LMMs. The experimental results show a significant improvement, with comparisons conducted across various tasks and datasets." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a dual-modality guided prompt learning framework (ModalPrompt) tailored for multimodal continual learning to effectively leran new tasks while alleviating forgetting of previous knowledge. Extensive experiments demonstrate the superiority of the proposed method." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The proposed method lacks substantial novelty, as prompt learning has already been widely used in fine-tuning pre-trained vision-language models in the continual learning setting.\n2. The baseline is too weak, thus the effectiveness of the method is not very convincing. For example, the baseline accuracy of zero-shot on the REC task is 0.00." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "- I will ask the authors to revise the entire paper to clarify their method and arguments.\n- In the main text the authors repeatedly emphasize that their method is time-efficient in the sense that the time complexity of inference depends on the number of selected prompts rather than tasks. However, I find this unclear. First, during the inference for each task sample, one needs to compute the similarity with all the prompts, whose number equals to the number of tasks. If we disregard such selection computation, why should other methods exhibit an $O(N_{task})$ time complexity?\n- To illustrate the importance of the dual-modality guidance, the authors compared the full results with those from using only image or text modalities. This comparison could be biased, as it relies solely on $\\alpha$ or $\\beta$ for prompts selection in the latter case. To ensure fairness, for example, one could use two different text encoders to obtain two estimates of text-based similarities $\\beta$ and $\\beta'$. This allows for a comparison of results using $\\alpha + \\beta$ with those using $\\beta + \\beta'$. Can you carry out this comparison and show the results?\n- There seems to be a discrepancy between results in Fig.5 and Fig.6: GQA task features show their highest similarity with ImageNet prototype features (Fig. 5). yet the selected prototype prompts are primarily from the GQA task (Fig. 6)." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "In large models like LLMs and LMMs, learned prompts serve as new \"viewpoints\" that enhance the performance of the underlying LMMs on specific tasks. I believe exploring prompt-based \"continued learning\" techniques can be practically beneficial, especially with the availability of powerful LMMs." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a continual learning scheme for LMMs based on prompt selection and fusion. Experiments on eight datasets show the effectiveness of the proposed method." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The paper is difficult to read, as it presents simple ideas in an abstract and complex manner. It requires a substantial revision before one can properly evaluate its soundness and contribution.Thus, I do not believe it is ready for publication at ICLR. Here are some areas of confusion I encountered:\n- In line 161, it states “The characteristic of LMM continual learning includes: …” It is unclear whether the authors refer to a general consensus on LMM continual learning or their specific proposals.\n- The summation in Eq.(3) lacks a dummy variable. Are you summing over individual prompts within a set for a specific task $t$?\n - Consider using $\\bar{x}$ for the average of prompts, as bold symbols could be confusing since they typically represent vectors.\n- In line 201, the projection should be defined as $\\text{Proj}_v(\\cdot):\\mathbb{R}^{d_v}\\rightarrow\\mathbb{R}^{d_t}$.\n- In Eq.(7), What is $X_p$? Is it the collection of all prompts? It's unclear how prompts are selected in your process.\n - One possible understanding: You have $N$ prompts for each of the $T$ tasks, so $T\\times N$ in total. The selection is performed over all the $T\\times N$ and produce $k$ most relevant ones.\n- Line 242 states, “To enhance knowledge transfer, the dual-modality features could serve as guiding cues for prompts to accurately get close to multimodal distributions of current task in feature space.” What are the dual-modality features? Are they the features of the current task? What do you mean by “multimodal distributions”? I don't think those terminologies are self-explanatory and commonly used in the field. Why is the closeness to the distribution helpful in enhancing knowledge transfer?\n- Eq.(9) abuses the symbol $\\mathbf{x}^t_p$ for prototype features, the same term is used for the “prompt features” in Eq.(3).\n- In Eq.(10) what are the definitions of $\\alpha^{\\le t}$ and $\\beta^{\\le t}$? What is the shape of $\\tilde{X}^t_p$?\n- In line 265, where do you define the parameters $\\theta_p^t$ of prototype prompts? \n- In Table 1, what is the metric of the first two methods?\n- In Table 2, what do $B_i$ and $M_i$ represent in the second row?\n- Previous text implies that “number of selection prompts $k$” refers to selecting the top-k most similar prompts. However, by line 448-455, it seems $k$ refers to the number of sets of prototype prompts. Which is the correct understanding?\n- Line 456 is confusing when it mentions “choosing three sets of prototype prompts.” Based on subsection 3.2 (line 237, “we term set of prompt for each task as prototype prompts”), shouldn’t the number of prototype prompt sets match the number of tasks, which is eight?\n- In Fig.5, it is not clear what quantity is plotted. Is it the average similarity between the prototype features and task features across all in task samples and targeting prototypes?\n\nIn addition, the visualization subsection at P.10 provides little information. Cherry-picking examples do not represent the overall behavior of your model. and I don't understand how these examples support the claim that your model retains previously learned knowledge." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1.\tThere are numerous methods for multimodal prompt learning. Did the authors explore other approaches, and if so, how effective were they? \n2.\tAdditionally, why does the baseline comparison only include the LoRA method? Are there other fine-tuning methods considered? Could a direct comparison between LoRA and prompt learning be potentially unfair? \n3.\tIs there any comparison of FPS, storage, and speed?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1.\tIntroduces an innovative, data-efficient solution to catastrophic forgetting, critical for LMM applications in dynamic task environments.\n2.\tDemonstrates strong empirical performance with improvements across key continual learning metrics.\n3.\tEfficient design enables lower computational cost, making it scalable for broader application." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents MODALPROMPT, a dual-modality guided prompt framework designed to address catastrophic forgetting in large multimodal models (LMMs) during continual learning. LMMs, which integrate visual and textual processing capabilities, encounter performance degradation when sequentially learning new tasks. To address this, MODALPROMPT leverages dual-modality (image and text) prompt learning to enable continual learning without task-specific expansion or data replay, which can be resource-intensive and raise privacy issues. By combining task-specific prototype prompts with a selection mechanism informed by image-text distributions, the model achieves improved task retention and transfer of knowledge across a variety of multimodal benchmarks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1.\tThe baseline lacks a comparison with other prompt learning methods.\n2.\tComplexity in configuring prompt numbers and selection features may limit broader accessibility without further simplification or automation.\n3.\tModalPrompt needs to convincingly differentiate itself from prior work in prompt-based continual learning, likely through robust comparative experiments and ablations." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "See weakness." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The question investigated in this paper is critical and significant in the current deep learning community.\n2. The paper proposes a novel prompt learning framework for rehearsal-free continual learning of LMMs.\n3. They conduct extensive experiments to demonstrate the effectiveness and inference speed of proposed methods." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper explores continual learning in Large Multimodal Models, focusing on the challenge of enabling models to continuously learn across sequential tasks. The authors critically assess the limitations of existing approaches and propose a novel dual-modality guided prompt learning framework for multimodal continual learning. Extensive experiments show that the proposed method significantly enhances both performance and inference speed." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Although the experiment improves the performance and inference speed, the proposed method involves modality-specific prompts for each task, which is too simple compared to existing work that devises advanced prompt strategies in visual scenarios. Simultaneously, they lack of comparison with the amount of prompt-based methods. Such as: DualPrompt [1], L2P[2], CODA-Prompt[3].\n2. There exist some typos in the paper:\n 1. in line 100, `prpredominant'.\n 2. in line 128, ... set `prompt of prompts' ... \n3. The author proposes the setting of refrains from computation expansion in proportion to the number of tasks. Whether means we can continuously learn the sequential data in one model and the performance will continuously improve. In other words, how many tasks can the proposed method effectively handle within one model?\n4. In the experiment, there is a lack of results that compare one task in the continuous process, i.e. compare the performance at the time axes, which directly reflects the transfer capability when more previous knowledge is learned.\n5. There is no difference between the two items in equation 12 with the add operation.\n6. How does the proposed method assess forgetting? Does it require saving a lightweight projection layer for each task, or should the projection layer from a previous task be re-tunned after learning a new one?\n7. In Line 203, why does the encoder of visual E_I and textual E_T in CLIP realize the mapping of E_I(·) : R^{n_v ×d_v} →R^{d_v} ,E_T(·) : R^{n_t ×d_t} →R^{d^t}, which should exist error description?\n\n[1]. DualPrompt: Complementary Prompting for Rehearsal-free Continual Learning\n\n[2]. Learning to Prompt for Continual Learning\n\n[3]. CODA-Prompt: COntinual Decomposed Attention-based Prompting for Rehearsal-Free Continual Learning" }, "withdrawal_confirmation": null }, { "TLDR": { "value": "a novel prompt learning framework for continual learning of large multimodal models" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024dualmodality,\ntitle={Dual-Modality Guided Prompt for Continual Learning of Large Multimodal Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=04TRw4pYSV},\nnote={under review}\n}" }, "abstract": { "value": "Large Multimodal Models (LMMs) exhibit remarkable multi-tasking ability by learning mixed datasets jointly. However, novel tasks would be encountered sequentially in dynamic world, and continually fine-tuning LMMs often leads to performance degrades. To handle the challenges of catastrophic forgetting, existing methods leverage data replay or model expansion, both of which are not specially developed for LMMs and have their inherent limitations. In this paper, we propose a novel dual-modality guided prompt learning framework (ModalPrompt) tailored for multimodal continual learning to effectively learn new tasks while alleviating forgetting of previous knowledge. Concretely, we learn prototype prompts for each task and exploit efficient prompt selection for task identifiers and prompt fusion for knowledge transfer based on image-text supervision. Extensive experiments demonstrate the superiority of our approach, e.g., ModalPrompt achieves +20% performance gain on LMMs continual learning benchmarks with x1.42 inference speed refraining from growing training cost in proportion to the number of tasks. The code will be made publically available." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Continual learning", "Large multimodal models", "Efficient learning", "Prompt learning" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/f557d72e5e53556a321bc19c7bcf56100cd4753a.pdf" }, "presentation": null, "primary_area": { "value": "transfer learning, meta learning, and lifelong learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Dual-Modality Guided Prompt for Continual Learning of Large Multimodal Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "- Did you find major differences between the two datasets in terms of atomizing claims in documents? It seems to me that this would be more structured in a medical Q&A dataset, as compared to LLM-user interactions." }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- Extremely well-written paper, with clear motivation and research questions.\n- The figures in the paper are very informative, and do an excellent job of conveying key information. The running examples were great for readability.\n- Clear problem statement, with adequate background and justification of design choices. Creative use of LLMs as an evaluation for text-coherence.\n- The results about access to different sets of auxiliary information were really interesting to me. The hypothesis about the non-uniformity of LLMs' instruction-following seems intuitive, but would be interesting to quantify this in its own right.\n- The human subject experiments were a nice touch - useful to know the capabilities of the two models in this context." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The manuscript seeks to highlight privacy concerns in text-sanitization techniques, by a) proposing a semantic-similarity based privacy metric for re-identification/matching attacks, and b) evaluating state-of-the-art defenses against such inference under the proposed metric. \n\nThe authors use a 2-step approach; in the first 'linking' step, sanitized documents are compared to externally known auxiliary information about a target individual using a TFIDF-based sparse retriever, and in the second 'semantic matching' step, a language model is used to assess similarity between atomic claims in the retrieved document and those in the original, unsanitized document.\n\nThe paper then evaluates several defense strategies to quantify information leakage under the above framework, and find that DP based methods may provide some of the strongest protections, albeit at the cost of data utility." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Can't think of any immediate flaws or weaknesses. Happy to discuss further once other reviews are in." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "L081, L099: Just to be sure: If I understood correctly, \"claims\" can refer to any sensitive or non-sensitive information in the original texts?\n\nL101: If (1) PII removal had 94% leakage (inferable claims), and (2) data synthesis (with or without DP?) has 9% lower leakage (i.e., 85% ?), why does (3) data synthesis without DP state 57% << 85% leakage?\n\nSection 2.3 Linking Method:\n- L146: Could you briefly motivate the use of the sparse BM25 retriever? Did you consider dense retrieval methods (say, using some form of text embeddings)? What are the benefits of BM25 vs. other sparse methods, or of sparse methods vs. dense methods, in particular for the linkage task at hand?\n- L147-148: You state your \"approach aggregates the auxiliary information into a single text chunk\" -- Does that mean you combine (concatenate?) _all_ \"atomized claims\" x^(i)_j across all j into the \"aux info\" {\\tilde x}^(i)? (Wouldn't hurt to write this down more explicitly.) (Just found the info in L296/Section 3.3 that the attacker gets 3 random claims for the linkage phase. I find that a bit late, it would be better to mention it directly in Section 2.3 to avoid confusion/guessing on the side of the reader.)\n\nSection 2.4 Similarity Metric:\n- L156: Can you give a specific reason for querying only with claims that were _not_ utilized in the linking phase? Besides, how do I know whether a claim about an original document was used for linking? If _all_ atomized claims are combined into the aux info (cf. previous question), and the aux info is used as query in the retriever, wouldn't this imply that all claims are already consumed in the linking phase?\n- L159: If I understood correctly, you define the \"similarity metric\" µ between the original and linked documents by querying a language model to judge the document similarity, where you assign values on a scale from 1 (for 'identical documents') to 3. I wonder if it would make more sense to start the scale at 0, since mathematically, a metric has the property of evaluating to 0 for identical inputs. (In your case, we would get \"µ(x,x) > 0\" instead of \"µ(x,x) = 0\".)\n- How do I know that the atomized claims, which are used to compute µ and hence to measure privacy preservation, are actually privacy-relevant, and not just some arbitrary, privacy-*in*sensitive facts?\n\nL313: I'm confused regarding the symbol \"µ\" seemingly used for multiple purposes. It is defined in Sec. 2.4, but here, you also use it for another metric induced by ROGUE-L scores.\n\nL317 (also L386): You state \"zero-shot prompting achieves an accuracy of 0.44\" for MedQA task utility, but why am I unable to find that result in Table 1 (for \"No Sanitization\", it says 0.69)?\n\nTable 1:\n- Calling the privacy metrics \"Overlap\" and \"Similarity\" is very confusing, since they actually mean the opposite (high lexical overlap and semantic similarity would indicate a high agreement between the two documents, but high scores in Table 1 mean good privacy). Name them lexical/semantic \"distance\" instead?\n- Talking about metrics: In Equation 1 you define a \"privacy metric\", I guess that is what is reported under the (why differently named?) \"Semantic Similarity\" column in Table 1. It is based on the \"similarity metric\" from Section 2.4, which has values between 1 and 3 -- How does it end up with values between 0 and 1 in Table 1?? I couldn't see any discussion on some form of normalization of these scores. The expected value of scores >= 1 in Eq. 1 would still result in a value >= 1, and not in [0,1). Please double-check how you actually compute the metrics. Try *not* to distribute information pertaining to one concept across the paper, but put it concisely into one place if possible. Also prefer consistent naming.\n\nTable 2:\n- The effect of \\epsilon appears surprisingly small to me, with only minimal changes across all metrics even when comparing \\epsilon=3 and 1024. Can you explain this behavior?\n- It would be interesting to compare with a random baseline where the utility is determined from completely random texts -- to rule out that 0.4x task utility in the case of MedQA can already be achieved based on completely random input (say, if the dataset suffers from strong class imbalance and the classifier always just guesses the largest class, thus obtaining an overly optimistic accuracy).\n\nTable 3:\n- L404: What exactly is the \"linkage rate\"? Please specify.\n- L423: Contradicting statements: Here, you state the last three claims are used, previously in L296, you mentioned 3 randomly selected claims.\n\nL443/Section 2.4: If you can switch the similarity metric µ also to ROGUE-L, please already state this as possible option in Section 2.4 where you introduce µ. Currently, you only say there that µ is determined by querying a language model.\n\nLastly, what are your thoughts on information that is both privacy-sensitive and utility-relevant, say, if one or more atomized claims are also strongly related to a downstream task? For instance, what if an atomized claim turns out to be \"John likes baseball\", and one of the WildChat categories is \"baseball\", too? Feel free to substitute \"baseball\" with something more delicate, such as \"drinking alcohol\".\n(Case A: If the baseball aspect is kept in the sanitized document, both µ and the chi^2 distance should be small, indicating poor privacy but good utility. Case B: If the baseball aspect was redacted, both µ and chi^2 should be larger, indicating better privacy but poorer utility.)\n\nAdditional considerations for related work:\n[1] also highlights the insufficiencies of superficial sanitization methods for text. [1] and also [3,4,5] propose differentially private methods that obfuscate texts. An evaluation framework for text rewriting has also been introduced previously [6].\n[2] has been published in parallel with (Yue et al., 2023) and also suggests differentially private synthetic text generation.\n- [1] Weggenmann & Kerschbaum, \"SynTF: Synthetic and Differentially Private Term Frequency Vectors for Privacy-Preserving Text Mining\", SIGIR 2018\n- [2] Mattern et al., \"Differentially Private Language Models for Secure Data Sharing\", EMNLP 2022\n- [3] Weggenmann et al. \"DP-VAE: Human-Readable Text Anonymization for Online Reviews with Differentially Private Variational Autoencoders\", WWW 2022\n- [4] Igamberdiev & Habernal, \"DP-BART for Privatized Text Rewriting under Local Differential Privacy\", ACL Findings 2023\n- [5] Bo et al., \"ER-AE: Differentially Private Text Generation for Authorship Anonymization\", NAACL 2019\n- [6] Igamberdiev et al. \"DP-Rewrite: Towards Reproducibility and Transparency in Differentially Private Text Rewriting\", COLING 2022" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "Formalizing linkage attacks for unstructured text data is a nice and useful contribution, and enables a systematic evaluation of various (novel) text sanitization methods in the future.\n\nWhile not entirely new, cf. e.g., [1] and the already cited (Stadler et al., 2022), the observations that superficial sanitization methods (such as PII removal) are often insufficient to properly protect privacy remains important.\n\nFor most parts, the paper is well written and easy to follow. However, there are some uncertainties about metrics and inconsistencies between numbers reported in the texts and tables, which are (in my view) confusing to the reader and undermine the validity of the currently reported results." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes a framework to evaluate sanitization methods for releasing datasets with textual data. They highlight that obvious methods such as removing explicit identifiers like names is insufficient for properly protecting privacy, since other semantic details can also leak private information. Also, auxiliary information about an individual may be linkable to a supposedly sanitized target document, thus allowing an attacker to infer or recover further sensitive details.\n\nThe goal of the framework is the quantification of information leakage from sanitized documents given auxiliary information about the document's owner or author.\nThe framework proposes to determine auxiliary information from each original documents by extracting individual \"claims\". For each document, the attacker is given a subset of claims and runs a sparse retriever to find the best-matching document from the set of sanitized documents.\nThey then define a similarity metric, which is either determined by an LLM or the ROGUE-L score, to compute the similarity between the retrieved document and the remaining claims extracted from the original document.\nAdditionally, they define task-specific utility metrics for each evaluated dataset.\n\nIn the evaluation, the authors consider two datasets: MedQA from the medical domain with a question-answering task, as well as WildChat consisting of online conversations with ChatGPT and a text categorization task. They also consider a range of sanitization methods that either work by removing PII or by generating synthetic data, the latter also with the option of providing differential privacy.\nIn each scenario, the newly introduced semantic and lexical privacy metrics are computed, along with task-specific utility measures as well as the quality (coherence) of the sanitized texts. Lastly, they perform a human evaluation to determine which variant of the privacy metric best matches human preferences." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I stumbled across some inconsistencies between the numbers reported in the Tables and discussed in the text. Please double-check (cf. questions) and update, or explain the differences.\n\nSome details about the metrics and their computation remain unclear (cf. questions). Please try to use consistent naming and define concepts (such as the definition of metrics/distances) in one concise and consecutive piece of text (not spread across several sections).\n\nL323: I think the conclusion from a \"disparity between lexical and semantic similarity\" to \"these techniques primarily modify and paraphrase text without effectively disrupting the underlying connected features and attributes\" is made a bit prematurely: Both are entirely different measures, and even for \"no sanitization\", the lexical score is twice the semantic score. Also, what would happen if you shifted the (apart from the ordering: somewhat arbitrarily) assigned scores for the \"similarity metric\" in Section 2.4 from {1,2,3} to {0,1,2} or to {1,10,100}?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "N/A" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper focuses on the privacy leakage of text data. The authors design different prompts for LLM to evaluate the semantic similarity between two sentences, which is interesting. The experimental results are extensive. Multiple data sanitization techniques are included in the evaluation framework." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces a privacy evaluation framework for data sanitization methods, specifically data anonymization and data synthesis, in the context of natural language. The framework can be summarized as follows: first, random records from the original data are sampled as the auxiliary data; then, an information retrieval technique is used to link the auxiliary data with the sanitized data, and an LLM is utilized to evaluate the semantic similarity between the original records and the linked records. The final similarity scores denote the degree of privacy leakage of the sanitized data." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The main issue of this paper is the definition of privacy leakage, which the authors equate with semantic similarity between auxiliary data and sanitized data. However, the semantic information of a sentence reflects its utility. If the semantic content of a sanitized sentence is altered, the sanitization method would be useless. Traditional data anonymization methods aim to remove only identifiers from a data record rather than all information. In this context, identifiers should be the privacy focus, and privacy leakage should refer specifically to identifier leakage.\n\n2. The technical novelty is relatively limited. The linking step uses the existing BM25 retriever, while the semantic similarity evaluation mainly relies on established prompt engineering techniques.\n\n3. The findings are not particularly interesting, as it is well-known that simple data anonymization and data synthesis techniques are insufficient to protect data privacy. This paper's findings merely confirm that this limitation also applies to text data.\n\n4. The numerical results rely on LLM output, which is relatively qualitative and less persuasive. Additionally, querying LLaMA three times for consistency seems unnecessary; disabling sampling parameters in text generation should ensure consistent results from LLaMA for the same query." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please refer to my weaknesses. Also, I have a few new questions.\n\n1) How can your method extend to other datasets? Is there any real auxiliary data that can be used instead of creating overlapped auxiliary data from the original records?\n\n2) Regarding the concept of privacy, is converting the age of 23 to early 20s a privacy breach? Such conversion is commonly adopted by K-anonymity." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The paper is well-written and easy to follow. All the included sanitization methods are up-to-date and well-explained.\n\n2. The proposed method is straightforward in decomposing the re-identification with linking and matching methods. \n\n3. Experimental results are comprehensive with sufficient ablation experiments. The included baselines are solid and up-to-date." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper investigates the current limitations of existing textual data sanitization methods. By considering re-identification attacks with known auxiliary information, the paper shows that a sparse retriever can link sanitized records with target individuals even though the PII patterns are anonymized. Instead, the paper proposes a new privacy evaluation framework for the release of sanitized textual datasets. The paper considers two datasets, including MedQA and WildChat, to show that seemingly innocuous auxiliary information can be used to deduce personal attributes like age or substance use history from the synthesized dataset. Experimental results also verify that current data sanitization methods create a false sense of privacy only on the surface level." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. My major concern is that the auxiliary data is highly contrived. Based on my understanding, each auxiliary sample is the subset of exact atomics from the target record. For example, in Fig 1, Auxiliary Information contains two atoms of the Original Record. That is, if you only consider de-identified, sanitized records, it is very easy for your BM25 retriever to get the sanitized target. In real-world re-identification attacks, there is no such auxiliary information that has many exact overlapped n-grams as original records. \n\n2. For the claim that 'private information can persist in sanitized records at a semantic level, even in synthetic data,' if you consider DP generation, the privacy level is indicated by $(\\epsilon, \\delta)$. That is, your linked record may not be the original target sample. DP introduces random noise to offer plausible deniability and protect the original record's privacy.\n\n3. The implemented methods for the proposed privacy evaluation framework only integrate various existing components by using the contrived auxiliary data. It is not likely to scale this framework for a large number of overlapped atoms." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "Privacy evaluation for quantifying disclosure risks of sanitized dataset release beyond surface level, exposing false sense of privacy" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024a,\ntitle={A False Sense of Privacy: Evaluating Textual Data Sanitization Beyond Surface-level Privacy Leakage},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=04c5uWq9SA},\nnote={under review}\n}" }, "abstract": { "value": "The release of sensitive data often relies on synthetic data generation and Personally Identifiable Information~(PII) removal, with an inherent assumption that these techniques ensure privacy. However, the effectiveness of sanitization methods for text datasets has not been thoroughly evaluated. To address this critical gap, we propose the first privacy evaluation framework for the release of sanitized textual datasets. In our framework, a sparse retriever initially links sanitized records with target individuals based on known auxiliary information. Subsequently, semantic matching quantifies the extent of additional information that can be inferred about these individuals from the matched records. We apply our framework to two datasets: MedQA, containing medical records, and WildChat, comprising individual conversations with ChatGPT. Our results demonstrate that seemingly innocuous auxiliary information, such as specific speech patterns, can be used to deduce personal attributes like age or substance use history from the synthesized dataset.\nWe show that private information can persist in sanitized records at a semantic level, even in synthetic data. Our findings highlight that current data sanitization methods create a false sense of privacy by making only surface-level textual manipulations. This underscores the urgent need for more robust protection methods that address semantic-level information leakage." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Privacy", "NLP", "Text", "Reidentification", "Data Release", "Sanitization", "Anonymization" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/b617ee9754da2d62065cb9dc97b3024b9e28590d.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "A False Sense of Privacy: Evaluating Textual Data Sanitization Beyond Surface-level Privacy Leakage" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "I would think similar studies are done for other fields such as electric vehicles. Are there better regulations for reporting in those? What are some other parallels and what other gaps can we find in transparency compared to them?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper raises very important concerns about transparency in the area of energy and water consumption required for developing as well as using LLMs.\n2. This paper includes aspects of this process which have not been reported before such as hyper-parameter tuning and manufacturing of GPUs.\n3. The discussion around these calculations is useful for others to understand the environmental implications of doing AI research." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper provides estimates and insights into power and water consumption during training and inference of Large Language Models (LLMs). Many of these estimates are based on real experiments in training these models of different parameter sizes. Some are rough estimates for cases where experiments were not possible, such as GPU manufacturing. The paper highlights that there are side activities such as data center cooling, development of the model etc. which are not accounted for in the literature reporting carbon emissions from model training." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. While some of the calculations are clear and seem reproducible as long as some of the manufacturer specific quantities are known, I am not 100% certain if all the steps can be followed by others. It would be useful if the authors can confirm whether someone can apply their methodology for similar experiments/calculations and if the paper contains all the details needed to do so.\n2. It would be helpful if the `Development` part of Section 4.1 can provide more details of what it covers i.e. what kind of HPO, what was the search space, what scaling experiments etc.\n\nMinor:\nmore transparency from developers on when, where, and how they a;re training their models -> more transparency from developers on when, where, and how they are training their models" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "I would like the authors to address W1 and W2, whether they agree or not, and if they do, how they plan to address them." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "S1: This paper provides the first-ever comprehensive end-to-end view of the environmental impact of an LLM. It is highly valuable to have this data published, as we, as a research community, need to have a complete view of how current training and deployment practices affect CO2 production and water usage. The insights and data provided can be used as a building block to argue future performance improvements not only for $ cost reasons but also to quantify their environmental impact. \n\nS2: The authors take care to question current naive assumptions like a 100% power draw during training, making this paper stand out. While this is a low bar, research on environmental impact in LLM training has had its share of invalidated research due to these minor, overlooked details.\n\nS3: The authors estimate the water usage during the development of an LLM, which I have not seen before in this line of research. This adds a new dimension to the environmental impact, providing a more complete picture of how current AI practices affect our environment." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper estimates the real-world environmental impact of training LLMs, including hardware manufacturing, model development, final training runs, and deployment. They vary model sizes, model architectures, and training time, providing a first view of the embodied CO2 production and water usage from AI systems at these scales. The entire study results in a production of 270t CO2 and a usage of 1.1M liters of water.\n\nAdditionally, it finds that model development accounts for 80% of the environmental impact, which is often not stated in related work. Additionally, they find that power draw has highs and lows due to checkpointing, creating a mixed load on the electrical grid, which may not easily be addressed, resulting in control challenges for power providers." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "W1: The result of model development being a large chunk of the environmental impact is not too surprising, but I agree that it is important to track and present in this paper. I am wondering about the representativeness of the data presented in this paper for model development and whether we will see a similar trend continue in the future. Given that this is a key contribution outlined in the abstract, I question whether the number of 80% will change significantly in future related work and if there are steps to take to present this more confidently. I am afraid that researchers in related fields take the final training costs and multiply them by 5x due to the results in this paper.\n\nW2: The second point of discussion has a similar issue as W1. While I agree that oscillating power draw may be a problem for power providers, I hesitate to agree that this is an issue at large. GPU-memory checkpointing has been shown to be possible by Gemini (https://arxiv.org/pdf/2312.11805), which likely reduces this time to sub-seconds. I am not against keeping this insight in general, and that power draw may be an issue for future techniques, but I question the future-proofness of this discussion point. Also, this being a problem for power providers could be explained in more detail and what this implies for the environmental impact.\n\nW3 (minor issues):\n* The EU AI Act could be included in 5.1 as it also includes the environmental impact of AI systems (e.g. Art 95 https://artificialintelligenceact.eu/article/95/)\n* Figure 1 takes a lot of space for the amount of information it provides. A double y-axis may not be the best visualization as it makes it harder to initially grasp the information. Maybe using two scatter plots would make the visualization more compact and easier to understand?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. In table 2, why is OLMo reported twice is there a difference?\n2. I am curious to know how much time it took to train your models. Given the hardware resources you had (up to 64 HGX servers with h100), why was your study limited to 7 billion parameter models?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The discussion of energy consumption and costs incurred by training and deploying LLMs is a rising and highly relevant topic.\nThis paper has the following strengths:\n\n1. It is one of the first studies to report the environmental impact of model development, not just the final training runs, highlighting the significant but often overlooked costs associated with hyperparameter tuning and other development activities.\n2. Cost evaluation encompasses power consumption, carbon emissions, and water usage and is not confined to a single aspect.\n3. Reporting the costs and putting them into perspective with real-world equivalencies." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper discusses the impact of large language models on the environment by studying the costs associated with training and deploying them. It explores the hidden cost of training a model, particularly when it comes to hardware manufacturing and the pre-training steps of creating a model, along with the training and deploying costs. They run their experiments on a small set of models with parameter sizes ranging between 20 million and 7 billion parameters. The results show that models released 270 metric tons of carbon emissions and 1.137 million liters of water. Additionally, the author discusses the power fluctuation during training that is a result of model checkpointing." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Even though the topic is highly interesting he are some limitations to the paper:\n1. Lack of novelty: The issue of power consumption in LLMs has been widely studied, and this paper doesn't provide any additional ideas, metrics, or insights expect for the study development cost.\n2. The findings are based on a specific set of small models, which may limit the generalizability of the results to other models and data centers with different configurations and efficiencies\n3. The study does not include data from actual deployment and usage of the models, relying instead on simulated scenarios, which may not fully reflect the actual environmental costs. In fact, the paper has a limited set of inference simulations with very simplistic assumptions, which may not fully capture the real-world deployment scenarios and their environmental impacts. \n4. Some of the calculations rely on assumptions and estimates, particularly regarding the embodied emissions and water consumption of hardware manufacturing, which may not be entirely accurate.\n5. Limited comparison across models: The authors seem to have taken the carbon consumption of llama and OLMo in Table 2 from previous works without replicating results, which meant no water usage comparison for training. For deployment, they only compare with Llama.\n6. Given the small sizes of the models, the paper lacks an analysis of how their results scale to larger models." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "Measuring the environmental impact, including carbon emissions and water usage, from training a series of language models." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024holistically,\ntitle={Holistically Evaluating the Environmental Impact of Creating Language Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=04qx93Viwj},\nnote={under review}\n}" }, "abstract": { "value": "As the performance of artificial intelligence systems has dramatically increased, so too has the environmental impact of creating these systems. While many model developers release estimates of the power consumption and carbon emissions from the final training runs for their latest models, there is comparatively little transparency into the impact of model development, hardware manufacturing, and total water usage throughout. In this work, we estimate the real-world environmental impact of developing a series of language models, ranging from 20 million to 7 billion active parameters, trained on up to 5 trillion tokens each. When accounting for hardware manufacturing, model development, and our final training runs, we find that our series of models released $\\textbf{270 metric tons}$ of carbon emissions, equivalent to powering about 53 homes in the United States for one year, and consumed $\\textbf{1.137 million liters of water}$, equivalent to about 10 years of water usage by a person in the United States, even though our data center is extremely water-efficient. We measure and report the environmental impact of our model development; to the best of our knowledge we are the first to do so for LLMs, and we find that model development, the impact of which is generally not disclosed by most model developers, amounted to $\\sim$$\\textbf{80}$% of that of training. By looking at detailed time series data for power consumption, we also find that power usage throughout training is not consistent, fluctuating between $\\sim$15% and $\\sim$85% of our hardware's maximum power draw, with negative implications for grid-scale planning as demand continues to grow. We close with a discussion on the continued difficulty of estimating the environmental impact of AI systems, and key takeaways for model developers and the public at large." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "machine learning", "artificial intelligence", "language model", "large language models", "environmental impact", "carbon emissions", "water usage" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/09f57c232e7b56840cd95b200d0e74eeea7e5db9.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Holistically Evaluating the Environmental Impact of Creating Language Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Does splitting video into frames for CLIP’s visual encoder lead to a loss of spatiotemporal information, and wouldn’t a video encoder like Video Swin Transformer [2] better capture temporal dynamics?\n[2] Liu, Z., Ning, J., Cao, Y., Wei, Y., Zhang, Z., Lin, S., & Hu, H. (2022). Video swin transformer. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 3202-3211).\n2. How does LLaVA-Surg perform compared to other state-of-the-art multimodal methods? In addition to general multimodal models, a detailed comparison with models like ReAct would provide a more comprehensive evaluation. Has comparison with other two-stage methods [3] in VQA task been overlooked?\n[3] Gai, X., Zhou, C., Liu, J., Feng, Y., Wu, J., & Liu, Z. (2024). MedThink: Explaining Medical Visual Question Answering via Multimodal Decision-Making Rationale. arXiv preprint arXiv:2404.12372.\n3. Is the two-stage question-answer generation process applicable to other medical fields, and if so, what adjustments would be required? Additionally, validating the method’s performance on public datasets like RAD, SLAKE, and PathVQA would strengthen its generalizability." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1.\tThe authors provide a novel dataset, Surg-QA, which is a significant resource for training multimodal surgical models, covering diverse surgical procedures and question-answer pairs.\n2.\tThe two-stage pipeline for question-answer generation mitigates hallucinations in LLM outputs, resulting in higher quality and reliability of generated data.\n3.\tLLaVA-Surg demonstrates notable improvements over general multimodal models in zero-shot surgical video question-answering tasks, showcasing its efficacy in understanding surgical context." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces LLaVA-Surg, a multimodal large language model designed as a conversational assistant for surgical applications. To support this, the authors developed Surg-QA, a large-scale dataset containing 102,000 surgical video-instruction pairs, generated through a structured two-stage question-answer pipeline. This pipeline helps extract structured knowledge from surgical lecture videos, enabling the LLaVA-Surg model to understand complex surgical procedures and answer open-ended questions in a zero-shot setting. The model leverages CLIP for visual encoding and is fine-tuned on Surg-QA to specialize in surgical video question-answering, achieving superior performance compared to existing general-domain models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The paper should compare its model with recent multimodal LLM approaches, specifically ReAct (Yao et al., 2023), which combines reasoning and action for complex tasks.\n[1] Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023, January). ReAct: Synergizing Reasoning and Acting in Language Models. In International Conference on Learning Representations (ICLR).\n2. Using CLIP for frame-by-frame encoding lacks temporal modeling and increases processing costs and redundancy, burdening the LLM as frame count grows.\n3. The paper lacks an in-depth error analysis, especially regarding potential hallucinations or misunderstandings in complex surgical scenarios. Although the authors claim to reduce hallucinations, achieving perfect performance seems challenging.\n4. The model’s adaptability to other medical or clinical fields is unclear, as broader evaluations on datasets like RAD, SLAKE, and PathVQA are missing, which may limit its wider applicability." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "I am wondering about the WebSurg's policies on using their videos to train deep learning models, but I could not find any information about this in their terms of use." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "The paper is very well written and addresses its objectives. It also supports its claims and provides adequate experiments. Therefore, I am leaning toward accepting this paper, but I have some minor concerns regarding the legality of using WebSurg's surgical videos. I also have some questions:\n1. The authors mention that the model is limited by hallucinations, which is a serious concern for a surgical chatbot. Could you please provide more details, and types of hallucinations, and give some examples?\n2. Would it be possible to evaluate LLaVA-Surg on the SSG-VQA dataset? I am interested in knowing more about the breadth of your dataset and if it contains enough information for cross-dataset generalization." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- **Clarity**: The paper is well-written and easy to follow. \n- **Contributions**: This work makes a significant contribution to the development of surgical chat assistants. The dataset contains a wider range of surgical QAs compared to previous works. The proposed model and dataset may be valuable resources for researchers in this area." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces LLaVA-Surg, a multimodal conversational assistant based on surgical videos. Additionally, they introduce a new dataset with 102,000 question-answer pairs for training multimodal LLMs. The authors provide details of their data generation procedure, which is carefully designed to avoid hallucinations. The paper provides detailed comparisons with existing general-purpose and surgical-purpose datasets. Lastly, the authors provide a human and LLM evaluation of the dataset, showing consistent scores." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- **Dataset Availability**: The surgical videos are available on WebSurg and are not a contribution of the authors. Therefore, the data availability may be subject to license changes from the content owners and WebSurg.\n- **Hallucinations and Data Quality**: As the authors mentioned, there may be hallucinations in the dataset, since it is automatically generated. The authors provide chatGPT and human evaluations, but that is not enough to infer the data quality.\n- **Model Availability**: It is not possible to reproduce the results since the model is not available yet, but enough details are provided to support the paper." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "Since doctors are hired to do the annotation, have the possible ethical risks been resolved? For example, IRB approval, etc." }, "flag_for_ethics_review": { "value": [ "Yes, Responsible research practice (e.g., human subjects, data release)" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Improvement of the methodology.\n- Detailed Comparison." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The Surg-QA dataset, along with the two-stage pipeline, is a significant contribution to medical AI. \n- LLaVA-Surg’s ability to process and interpret surgical video content sets it apart from other models focused primarily on static images.\n- The language is clearly presented. The authors use precise and concise language so that the reader can easily understand the dataset, methodology, and results of the study." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a novel vision-language model, LLaVA-Surg, designed to assist in surgical settings. Leveraging the newly created Surg-QA dataset with 102K surgical video-instruction pairs, the model provides conversational, open-ended responses to questions about surgical procedures. Evaluations demonstrate LLaVA-Surg’s superior performance in surgical video question-answering, indicating its potential as a reliable tool in surgery-related applications." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Although the dataset is valuable, this storyline and methodology is too similar with LLaVA-Med [1]. Maybe the authors could think of improvements of this simple fine-tuning method (i.e., SFT) to make better use of this dataset.\n- The paper lacks comparative results. The current comparative models are rarely trained on surgical scene data, which is unfair. It is necessary to compare with a specific model.\n- Since doctors are hired to do the annotation, have the possible ethical risks been resolved? For example, IRB approval, etc.\n\n[1] Li C, Wong C, Zhang S, et al. Llava-med: Training a large language-and-vision assistant for biomedicine in one day[J]. Advances in Neural Information Processing Systems, 2023." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "Potential copyright problem for online data." }, "flag_for_ethics_review": { "value": [ "Yes, Legal compliance (e.g., GDPR, copyright, terms of use)" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. How can the quality of the data be ensured? The data collected may already contain a lot of noise and has been reprocessed by an LLM. Is there any person or clinician reviewing these raw data?\n2. Can the data be released? Are there privacy and permission risks associated with the collected data?\n3. The authors need to conduct more zero-shot evaluations on downstream tasks relevant to the surgical field, such as phase recognition, action/instrument classification, and other surgical domain VQA data to demonstrate the clinical usability of their method.\n4. The authors need to compare with more state-of-the-art methods. The comparison methods in Table 3 were all first released in 2023.\n5. The authors may verify their dataset on more benchmarks of SOTA Video MLLM architectures.\n6. Also, the authors need more zero-shot comparisons with the same VLM trained on other surgical datasets, to showcase the generalizability of their proposed dataset.\n7. The authors may evaluate the visual quality of the surgical videos themselves, as they are obtained from the website." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. With over 102,000 video-instruction pairs, this dataset is the largest in the surgical field.\n2. Structured data annotation pipeline using LLMs minimizes the risk of generating inaccurate or nonsensical content, improving dataset reliability.\n3. Releasing the dataset, model, and code publicly fosters further research and development in the surgical AI domain.\n4. The dataset can be a valuable resource for training and education, helping surgical trainees learn through interactive Q&A about real procedures." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a novel surgical multimodal dataset, which consists of over 102,000 video-instruction pairs generated through a two-stage pipeline, aimed at enhancing the understanding and conversational capabilities of surgical videos." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The paper does not address how the data's quality is maintained as the videos are obtained from the web. The clinicians have reviewed the output of their MLLM model, but the paper does not confirm whether clinicians or domain experts have reviewed the raw data to ensure accuracy and reliability.\n2. Concerns regarding the release, privacy, and permission risks associated with using sensitive surgical videos are not adequately discussed.\n3. The paper lacks comprehensive validation across essential surgical downstream tasks and other surgical QA datasets, which are crucial for demonstrating clinical usability. There is also a need for more rigorous benchmarking against a broader range of state-of-the-art video MLLM architectures to establish the dataset's utility and the model's performance more robustly.\n4. The comparison of the proposed methods with SOTA methods is limited and does not include the latest works. The manuscript also lacks evaluations with models trained on other surgical datasets, limiting the assessment of the proposed model's generalizability across different surgical scenarios.\n5. The paper may need to evaluate the visual quality of the surgical videos." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "Please address the weaknesses mentioned above." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "1. The pipeline is comprehensive: A two-stage question-answer generation process minimizes hallucinations by extracting information prior to generating pairs, which enhances data quality and reliability compared to Quilt-1M[1], which has a similar approach.\n\n2. Integrating surgical visual concept alignment data through action triplets improves text-visual alignment, enhancing the model’s grasp of surgical concepts.\n\n3. The idea is interesting: using the Spearman rank correlation between expert and GPT scores effectively validates the reliability of large-scale GPT evaluation.\n\n[1] Ikezogwo, Wisdom, et al. \"Quilt-1m: One million image-text pairs for histopathology.\" Advances in neural information processing systems 36 (2024)." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces LLaVA-Surg, the first multimodal surgical assistant capable of understanding surgical videos and engaging in open-ended conversations about them. The authors create Surg-QA, a dataset of 102,000 surgical video-instruction pairs, using a novel two-stage question-answer generation pipeline. This approach reduces LLM hallucinations and costs by breaking down the generation process. The resulting model demonstrates superior performance in surgical video question-answering compared to previous general-domain models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Could you provide results for the three existing surgical-domain datasets (EndoVis-18-VQA, Cholec80-VQA, and SSG-VQA) trained on Surg-QA? These results could demonstrate Surg-QA's potential as a foundational dataset in the surgical domain.\n\n2. Maybe considering to use other video VLM models, which provides a more sophisticated approach to temporal fusion than simple average pooling." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024llavasurg,\ntitle={{LL}a{VA}-Surg: Towards Multimodal Surgical Assistant via Structured Lecture Learning},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=063FuFYQQd},\nnote={under review}\n}" }, "abstract": { "value": "Multimodal large language models (LLMs) have achieved notable success across various domains, while research in the medical field has largely focused on unimodal images. Meanwhile, current general-domain multimodal models for videos still lack the capabilities to understand and engage in conversations about surgical videos. One major contributing factor is the absence of datasets in the surgical field. In this paper, we create a new dataset, Surg-QA, consisting of 102,000 surgical video-instruction pairs, the largest of its kind so far. To build such a dataset, we propose a novel two-stage question-answer generation pipeline with LLM to learn surgical knowledge in a structured manner from the publicly available surgical lecture videos. The pipeline breaks down the generation process into two stages to significantly reduce the task complexity, allowing us to use a more affordable, locally deployed open-source LLM than the premium paid LLM services. It also mitigates the risk of LLM hallucinations during question-answer generation, thereby enhancing the overall quality of the generated data. We further train LLaVA-Surg, a novel vision-language conversational assistant capable of answering open-ended questions about surgical videos, on this Surg-QA dataset, and conduct comprehensive evaluations on zero-shot surgical video question-answering tasks. We show that LLaVA-Surg significantly outperforms all previous general-domain models, demonstrating exceptional multimodal conversational skills in answering open-ended questions about surgical videos. We will release our code, model, and the instruction-tuning dataset." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Multimodal assistant", "surgical", "multimodal instruction-following data", "dataset" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/04d73daf100581d96e3a971dd358d0aad68ebdd1.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "LLaVA-Surg: Towards Multimodal Surgical Assistant via Structured Lecture Learning" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "1. 3D structures (conformers)\n\nAs mentioned in sec. 5.1 you use MMFF for molecular conformation generation.\n\na. Is it ETKDG geometry generation with further MMFF optimization?\nb. Since it is possible to generate several different conformers for a single molecular structure, did you assess the dependence of the model quality on the conformations? Is it necessary to optimize a generated with ETKDG conformer with MMFF?\n\n2. It would be reasonable to compare your approach for Zero-shot editing with conditional generation models for small molecules.\n\n3. Please, add experiments on the CHEBI-20 benchmark for the captioning task." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "New molecular multimodal LLM framework for simultaneous incorporation of 1d 2D and 3D representations.\nNew Transformer architecture MQ-Former." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The work proposes a novel multimodal LLM framework MV-CLAM for organic chemistry and MQ-Former — multi-querying transformer model for simultaneous 1D, 2D, and 3D molecular representation learning. Authors show SOTA results in two tasks of molecule-text retrieval and molecule captioning. In addition, authors claim that their approach allows zero-shot molecule editing and molecule-related question answering." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The claim of the state-of-the-art performance for molecule captioning is not satisfied, see the results in [6].\nThere is no comparison with the other strong retrieval methods for the molecule retrieval task, i.e. RAG.\nThere are various problems with the Zero-shot editing part of the paper. The task is not formally defined. There are no metrics nor baselines for it.\n\nThe QA part is practically absent in the paper, while claimed in the abstract and results parts..\nThere are many works on molecular conformation generation [1-4], it seems that SMILES and/or 2D-graph representation is enough for neural networks to reconstruct RDKIT conformations almost perfectly. It means that 3D input possibly does not add any new information to the model. There is no comparison of the 1D+2D+3D MQ-Former vs 1D+2D models in the paper.\n\nThere is no comparison with other works on multi-modal representation learning for molecules, e.g.: [5]. \n\n[1] Zhu, Jinhua, et al. \"Direct Molecular Conformation Generation.\"\n[2] Xu, Minkai, et al. \"GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation.\" International Conference on Learning Representations.\n[3] Jing, Bowen, et al. \"Torsional diffusion for molecular conformer generation.\" Advances in Neural Information Processing Systems 35 (2022): 24240-24253.\n[4] Lee, Danyeong, et al. \"Disco: Diffusion Schrödinger bridge for molecular conformer optimization.\" Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 38. No. 12. 2024.\n[5] Manolache, Andrei, Dragos Tantaru, and Mathias Niepert. \"MolMix: A Simple Yet Effective Baseline for Multimodal Molecular Representation Learning.\" arXiv preprint arXiv:2410.07981 (2024).\n[6] Liu, Zhiyuan, et al. \"ReactXT: Understanding Molecular\" Reaction-ship\" via Reaction-Contextualized Molecule-Text Pretraining.\" arXiv preprint arXiv:2405.14225 (2024)." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "See 'Weaknesses' section.\n1. Could the authors provide a more detailed explanation of the novelty of MV-CLAM compared to recent related work?\n2. Why was SELFIES not considered as a molecular modality in this work, given its advantages over SMILES in tokenization and alignment with LLMs?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper integrates both 2D and 3D molecular structures to enhance the model's understanding of molecular data.\n2. The paper includes detailed figures (Figure 1-3) that clearly explain the method's framework and training scheme. \n3. And the analysis of attention maps in Appendix A.4 provides valuable insights into the model's behavior." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a framework that leverages large language models (LLMs) to interpret and generate molecular captions. The work incorporates both 2D and 3D molecular structures to provide a more comprehensive understanding of molecules." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Compared to recent related work, such as 3D-MoLM (Li et al., 2024), the innovation in MV-CLAM appears incremental. While the paper claims to incorporate both 2D and 3D molecular structures for a more comprehensive understanding, the approach seems to merely extend the 3D-MoLM framework by introducing 2D components through MAT. The proposed MQ-former architecture does not demonstrate significant structural innovations beyond existing methods. A clearer articulation of the novel contributions and architectural advantages over 3D-MoLM would be necessary to establish the work's originality.\n2. The paper considers SMILES as an important molecular modality and notes that \"1D SMILES provide compact represen tation of molecular structures\", but does not mention SELFIES (Krenn et al., 2020) at all, which has been widely adopted in recent works due to its robust characteristics and tokenization-friendly nature. SELFIES offers inherent robustness and easier tokenization that aligns well with LLMs, making it a potentially more suitable choice for this application. \n3. Some images (e.g. the big image at page 18) are not vector graphics and lack titles or captions, which makes it confusing." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "* Add experimental comparison against more chemical language models on molecule captioning, e.g., nach0 [1], Text+Chem T5 [2], SciFive [3], PRESTO [4], GitMol [5].\n* For retrieval task (Table 1), is it possible to add chemical BERT-based encoders in addition to textual encoder SciBERT? (e.g., ChemBERTa)\n* Conduct additional experiments on other molecule captioning datasets such as Mol-Instructions [6] and CheBI20 [7].\n* For molecule-text retrieval, do you adopt a generative approach (e.g., GENRE [8]) or the task is formulated as a cross-modal embedding-based search by similarity (e.g., as in [9])?\n* In Figure 3, where does the textual description come from during prediction on a test set? As far as I understand the molecule captioning task, you are only given a SMILES string.\n* What is the LLaMA version you use? Add adopted HuggingFace checkpoints. \n* Even if you adopt a LLaMA with 7B parameters, MolT5 has less than 1B. Could not we just scale MolT5 to 3-5B parameters and obtain a better molecule captioning quality?\n* Why is MolT5 absent from the Table 1?\n* Add ablation study for SciBERT, 2D/3D molecule encoders, LLaMA2.\n* Add ablation study for training losses. For Molecule-text Contrasting loss, prove it requires two components. For Molecule-text Matching loss, explore the effect of negative samples.\n* Is it possible to generalize the methodology to unseen datasets and unseen SMILES? Given a SMILES, can I always obtain its 2D/3D representation and apply a pre-trained MV-CLAM model?\n\n\n\n\nTypos:\n* Line 102: transformer -> Transformer, Add reference.\n* Line 194: **$A$** under-specified.\n* Line 234: Missing citation for LoRA." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The description of the proposed methodology is easy to follow. The paper is well written in general.\n* The paper introduces a promising multi-view for approach for the infusion of specialized chemical knowledge into general-purpose pre-trained LLMs.\n* The proposed MV-CLAM achieves state-of-the-art on PubChem324K for molecule captioning and retrieval tasks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces MV-CLAM, a framework utilizing a novel multi-querying transformer (MQ-Former) to enhance the alignment of multi-modal molecular representations with text. By employing a shared self-attention layer, this approach effectively consolidates 2D and 3D molecular data into query tokens, improving performance in molecule-text retrieval and captioning tasks. Additionally, it demonstrates potential for zero-shot molecule editing and molecule-related question answering, thereby facilitating better characterization of chemical structures." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* The experimental evaluation of the proposed method is conducted on a single dataset for both task: molecule captioning and molecule-text retrieval.\n* The list of baseline models on molecule captioning only includes a single T5 language model while there are more recent works, including: nach0 and Text+ChemT5. \n* Some implementation decisions are not justified well enough. This includes: (i) the choice of SciBERT as a language encoder for MQ-Former; (ii) the choice of 2D and 3D encoders; (iii) introduction of $K$ query tokens instead of a single query token for each view; (iv) the choice of LLaMA2 as an LLM. It is unclear how the experimental results would change if each of the mentioned models is replaced with another one.\n* Incomplete ablation study. The necessity of (i) Molecule-text Contrasting and (ii) Molecule-text Matching losses is not proven experimentally. For (i), it is unclear whether two loss components required or the model will perform well with a single one. For (ii), the impact of negative sample is under-explored. \n* The effect of most hyper-parameters in the method's module on the resulting performance is understudied. For instance, query token count, negative sample count in MTM loss.\n* The methodology for molecule-text retrieval is unclear from the paper.\n* The applicability of the proposed methodology to broader list of datasets is questionable: it requires 2D/3D molecular data in addition to simple SMILES string representations." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "N/A" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- How does MQ-Former handle scenarios where 2D and 3D molecular information may not equally contribute to textual descriptions?\n- Could the authors include more molecular tasks, such as molecule generation or property prediction, to provide a more comprehensive evaluation of MQ-Former?\n- What impact does the weighting of the multi-objective training loss have on the model’s performance?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The paper aims to enhance cross-modal alignment by integrating 2D and 3D molecular views.\n- The model demonstrates improvements in molecule-text retrieval and captioning performance over baseline models.\n- The paper includes case studies and examples of zero-shot molecule editing." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes MQ-Former, an extension of the Q-Former framework, which incorporates a multi-query mechanism for aligning both 2D and 3D molecular data with textual information for enhanced molecule-text retrieval and molecule captioning." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The model lacks significant innovation, as MQ-Former primarily adds an extra branch to the existing Q-Former with only minor variations in training objectives.\n- Experiments are restricted to molecule-text retrieval and captioning on PubChem. The paper lacks essential molecular tasks like molecule generation and datasets like ChEBI-20.\n- The motivation for adding a branch to Q-Former, rather than simply using a 3D molecular encoder like prior works (e.g., 3D-MoLM), is unclear. \n- The paper’s presentation could be improved. Plots lack careful formatting, with text that is difficult to read due to small font sizes." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "MV-CLAM introduces MQ-Former, a model that aligns 2D and 3D molecular representations with text via a novel cross-modal projector, improving tasks like molecule-text retrieval, captioning, and question answering." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024mvclam,\ntitle={{MV}-{CLAM}: Multi-View Molecular Interpretation with Cross-Modal Projection via Language Model},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=06B23UkNid},\nnote={under review}\n}" }, "abstract": { "value": "Large language models (LLMs) have shown significant potential in the biomolecular domain, particularly by demonstrating that effective adaptation of molecular representations for LLMs can greatly improve the quality of molecular captions. Most previous works have focused on aligning unimodal molecular structures with text, overlooking the diversity of modalities. Naive approaches to aligning multi-modal molecular structures with text often lead to (1) separately aligned embeddings, (2) inconsistent textual representations, and (3) increased computational overhead. To address these challenges, we propose LLM framework MV-CLAM equipped with MQ-Former, a novel multi-querying transformer. This architecture introduces a cross-model projector facilitating the simultaneous alignment of 2D and 3D molecular representations to a unified text token. By employing a shared self-attention layer, MQ-Former preserves rich molecular embeddings across different dimensions while consolidating them into a universal molecular token. Our approach outperforms baseline models in both molecule-text retrieval and molecule captioning tasks. Additionally, our framework shows promising results for zero-shot molecule editing and molecule-related question answering. By effectively integrating multi-view molecular data into a format conducive to LLMs, our method serves as a valuable tool for enhancing the characterization and understanding of chemical structures, facilitating a more seamless transition from molecular data to textual descriptions. The source code of MV-CLAM is available in https://anonymous.4open.science/r/mv-clam-4827." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Molecule captioning", "large language models", "drug discovery", "molecule representation learning" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/96c1427eb118eda7038e3ca6308c3231d9a8911e.pdf" }, "presentation": null, "primary_area": { "value": "applications to computer vision, audio, language, and other modalities" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "MV-CLAM: Multi-View Molecular Interpretation with Cross-Modal Projection via Language Model" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* What is the rationale of choosing the Beta and Uniform distribution (beyond what is described in line 323-324). Are there any related works that you could cite to support this choice of distributions?\n\n* What is the scaling of complexity and cost (such as evaluation and communication) as the number of the agents increase? Are there any risks of agents colluding to achieve a suboptimal safety level?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "The paper is well-written and well-supported by both theoretical proofs and empirical results. It addresses the important area of AI regulatory via a multi-agent economic, game-theory type framework. There are a few assumptions to simplify the mechanism but they appear to be acceptable/realistic such as i) the regulator and the participating agents use data from the same distribution to evaluate and submit the safety level, and ii) safer models cost more to develop. The paper has help enhance the current AI regulatory work with a well-formulated framework and has a potential to have some significance in this domain." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a novel framework of auction-based regulatory mechanism as an asymmetric and incomplete all-pay auction. The mechanism is described mathematically and also shows good empirical results of enhancing safety and participation rates. The framework consists of a regulator and multiple participating agents. Overall, this is an interesting framework with good potentials to explore and create safer and more robust AI regulatory." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "There are no obvious weaknesses to note. While the paper is well-supported in the mathematical formulation and proofs, it perhaps could have provided more evidence on the experiments and empirical data. More description of how this framework can be applied in AI regulatory or in practice might help ground it further and make it relevant to a wider group of audiences." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "**Q1.** Is there a reasonable mechanism for estimating the market value ($v_i^d$) of a model before it is submitted to the regulator or even before the training phase begins?\n\n**Q2.** Considering that SIRA’s performance deteriorates at high safety thresholds, would a simple increase in the threshold serve as a better incentive in such cases, as it may more directly encourage safer model development?\n\n**Q3.** The authors mention that safety evaluations rely on IID assumptions for both agent and regulator data. How would the proposed mechanism adapt to non-IID settings, where the agent's training data might be maliciously poisoned, or where the regulator's evaluation data is collected through other means?\n\n**Q4.** Is the random comparison fair for all competitive agents? For example, if we have utility values such that $u_A > u_B > u_C > u_D$, and A and B are grouped together while C and D are grouped together, then B and D cannot receive the policy bonus. However, since $u_B > u_C$, this situation could be considered unfair to B." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "**Originality.** The approach presents a unique use of all-pay auction mechanisms in AI regulation, where each agent's utility is linked to model safety levels (training cost), model value (market returns), and premium (policy compensation), creating an incentive for improved safety compliance.\n\n**Quality.** The paper theoretically derives Nash Equilibria to back the proposed incentive structure, demonstrating that agents' rational behavior leads them to exceed the minimum safety threshold. The experimental results align with the theoretical model.\n\n**Clarity.** This paper is well-written and easy to follow. The authors provide clear descriptions of the auction-based model and detailed steps in the algorithmic design of SIRA, supported by both theoretical and empirical validation.\n\n**Significance.** This paper tries to tackle an essential issue in AI regulation by encouraging safer model deployment." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a new AI regulatory framework known as the Safety-Incentivized Regulatory Auction (SIRA), designed as an all-pay auction. SIRA aims to motivate model-building agents to prioritize safety beyond a minimum threshold by formulating the AI regulatory process as an asymmetric all-pay auction with incomplete information. In this framework, agents submit their models to a regulator, and those that meet or exceed a specified safety level become eligible for deployment and may also receive additional rewards based on the randomized pair comparison result. The authors theoretically prove that under some assumptions and when all agents adopt a certain strategy, the system reaches a Nash Equilibrium. Empirical results indicate that when safety threshold prices are in the middle (0.2~0.8), SIRA enhances safety compliance and agent participation by 20% and 15%, respectively compared with the basic regulatory method." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "**Rationality of the auction framework.** Considering the regulation process as an all-pay auction is not reasonable, at least in my opinion. Intuitively, safety-driven regulation establishes a minimum cost for the model-building agent. Every model-building agent must incur this cost, regardless of whether it can successfully meet the regulatory requirements. This represents an unavoidable exploration process within the model space. Even if we assume that all competitive agents know how to train their models to meet the safety threshold, accurately estimating the value of deployment remains a challenge. Thus, the framework may be overly simplistic in its approach to \"safety\" regulation.\n\n**Feasibility of Assumptions 1 and 2.** Assumption 1 fails when a model-building agent maliciously injects backdoor triggers into the model by altering the training dataset. Assumption 2 is also not straightforward. More cost (e.g., computational resources) does not necessarily equate to better safety. Safety also depends on other factors, such as the learning paradigm, model architecture, loss function design, and hyperparameter selection.\n\n**Performance at high thresholds.** As highlighted in the experiments, SIRA demonstrates limited advantages when safety thresholds approach the upper range (e.g., above 0.8), where its performance is similar to that of simpler reserve threshold models.\n\n1. Evan Hubinger, et al., Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training, arXiv:2401.05566." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "1. What is the technical challenge in the considered auction problem for AI models, compared to classic auction problems?\n\n2. Practical AI models are often very large. How can the safety of these model be evaluated? Given that the auction is done in a one shot setting, probably it is fine even if the model is large.\n\n3. I am more concerned about the compensation $v_i^p$, which needs to be provided by a regulator to implement the proposed auction algorithm. Why is this practical for existing AI models? How large does the compensation need to be? According to bidding equilibrium in Theorem 2, $v_i^p$ needs to be large for safer models. How could this be made up to compensate what the commercial AI models could achieve?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The topic considered in this paper is interesting and important. Regulations are needed to ensure AI safety.\n\n2. Theoretical results are provided whose proofs can be found in the appendix. I didn't check all the mathematical proofs.\n\n3. The paper is overall well-written and well motivated." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper addresses the challenges regulators face, particularly with the deployment of large language models that can amplify misinformation and societal division. It highlights the urgent need for effective regulatory frameworks to mitigate these risks and enhance user safety. Observing a gap in the availability of rigorous and realistic mathematical frameworks for AI regulation, the authors propose an innovative auction-based regulatory mechanism. This mechanism is designed to incentivize the development and deployment of safer AI models and encourage active participation in the regulatory process. It demonstrates through derived Nash Equilibria that the proposed auction mechanism effectively ensures that each participating agent’s optimal strategy aligns with submitting a model that exceeds a set minimum-safety threshold." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The way used by the paper to model the safety may not be realistic. It is assumed to be some safety level $s_i$ of a model $w_i$, which is expected to be less than $\\epsilon$. How is the safety measured for AI models using the metric mapping $S$ in practice? For common foundation models and LLMs, it might be hard to evaluate $S$ for $w_i$, especially given the size of $w_i$. What if a model provider take advantage of the inaccuracy of the safety evaluation to benefit itself?\n\n2. The proposed auction algorithm, together with the theoretical results and analysis seem quite standard. How does it differ from the classic all-pay auction results (for instance, Amann et al. 1996) in the setting for AI models? It is worth highlighting the technical novelty and emphasize why the proposed method is needed for AI models, given that it is claimed in Line 398-399 that \"To the best of our knowledge there are no other comparable mechanisms for safety regulation in AI.\"" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Beyond the questions listed in the Weakness section, here are some additional questions I have:\n- The framework assumes that the cost $M$ is the same across agents. This assumption seems unrealistic in practice, given that different agents may have varying models, training procedures, and resources, which makes the cost of aligning the safety levels different. If $M$ differs across agents, is there a way to adapt the framework to accommodate heterogeneous costs while maintaining its theoretical properties? \n- The paper didn't mention incentive compatibility, a key issue in auction literature. Is truthful report of $b_i$ guaranteed?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- A novel and important question, and strong motivation\n- Sound theoretical analysis\n- Genrally well-written" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors provides a formulation of the AI regulatory process as an all-pay auction, and design an auction-based regulatory mechanism that produces Nash Equilibria that induces safety considerations." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The authors' formulation of the regulatory process and safety components appears to be somewhat simplified and may diverge from current AI developments in a few key ways:\n- (Minor) The authors assume a fixed safety threshold, denoted as $\\epsilon$, for model development. While this may hold in domains such as drug approvals or medical equipment (as illustrated by the authors' N95 mask example), applying a similar framework to AI models is more challenging and complex.\n- (Minor) The model assumes that the test set used by regulators is drawn from the same distribution as the agent’s evaluation data. However, in the specific context of language models, techniques such as fine-tuning and reinforcement learning from human feedback (RLHF) can easily improve performance metrics if the evaluation distribution remains consistent. This weakens the argument that a single scalar value would sufficiently capture the intricacies of regulatory inspection.\n- The authors propose a strictly increasing relationship between safety and cost, arguing that \"safer models cost more to develop.\" However, they do not explicitly account for the trade-off between safety and the model's quality or usefulness in their framework. This omission raises questions, particularly since existing alignment approaches (e.g., RLHF) are often designed to balance helpfulness and harmlessness. In practice, a model could be made extremely safe (e.g., by providing only generic responses), but this could significantly reduce its usefulness without necessarily increasing development costs. In fact, under the authors' framework, one could submit a trivial model (e.g., one that always responds with \"Thank you for your input\"), bid the highest possible value, and meet the safety threshold $\\epsilon$ to claim the regulator's compensation. This suggests that achieving safety in some cases may not necessarily be costly unless the model’s quality or usefulness is held constant.\n- This issue could be exacerbated by the presence of open-source models like LLaMA, which may further incentivize the \"gaming\" of the regulatory system. Agents could enter the competition with low-cost variants of open-source models that prioritize safety at the expense of quality, primarily to secure the regulator’s compensation. Put it in a different way, low-quality models (which are safe but not useful) could flood the regulatory system, making it easier to claim compensation without delivering valuable AI products. This could distort incentives, where participants optimize for regulatory approval rather than producing high-quality, well-rounded models.\n\nFor the mechanism itself, a minor concern include the use of randomization, which introduces envy into the mechanism. With development costs potentially huge, this might lead to issues and discontent and distrust with the mechanism after the outcome is realized." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose an auction-based regulatory mechanism that incentivizes agents to develop and deploy safer AI models." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024auctionbased,\ntitle={Auction-Based Regulation for Artificial Intelligence},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=06GH83hDIv},\nnote={under review}\n}" }, "abstract": { "value": "In an era of \"moving fast and breaking things\", regulators have moved slowly to pick up the safety, bias, and legal pieces left in the wake of broken Artificial Intelligence (AI) deployment. Since AI models, such as large language models, are able to push misinformation and stoke division within our society, it is imperative for regulators to employ a framework that mitigates these dangers and ensures user safety. While there is much-warranted discussion about how to address the safety, bias, and legal woes of state-of-the-art AI models, the number of rigorous and realistic mathematical frameworks to regulate AI safety is lacking. We take on this challenge, proposing an auction-based regulatory mechanism that provably incentivizes model-building agents (i) to deploy safer models and (ii) to participate in the regulation process. We provably guarantee, via derived Nash Equilibria, that each participating agent's best strategy is to submit a model safer than a prescribed minimum-safety threshold. Empirical results show that our regulatory auction boosts safety and participation rates by 20% and 15% respectively, outperforming simple regulatory frameworks that merely enforce minimum safety standards." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Regulation", "Mechanisms", "Auctions", "Artificial Intelligence" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/712d0f5a1b8c06931f9d87edc2b1482e3b12a100.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/19ff95c79f6c6d7867316e5b6e85d1d4b88dec50.zip" }, "title": { "value": "Auction-Based Regulation for Artificial Intelligence" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "- The Qwen and Gemini model versions should be specified." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The method incorporates insights from a legal decision-making perspective, and provide two frameworks that simulate the human workflow in court." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents two multi-agent systems inspired by courtroom for evaluating the outputs of LLMs. The experiments show the proposed frameworks improve accuracy compared with a single LLM as a judge." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Lack of experiments: More evaluation datasets and baselines should be incorporated into the experiments. For example, LLM-based multi-agent evaluators such as PRD [1] and ChatEval [2] could be baselines. There are many datasets in this LLM-based evaluator topic, such as AlignBench [3], AUTO-J [4] and LLMEval [5].\n\n- The presentation needs to be refined: \n\n(a) The background (in both Section 1 and Section 2) is taking up too much space. This background can be concluded to make space for evaluation details in Appendix D. \n\n**(b) The Conclusion section and Appendix A.5 are likely to be AI-generated (according to GPTZero).**\n\n- The multi-agent systems will surely use more tokens compared to LLM-as-a-judge. What is the cost per run compared to other multi-agent frameworks (such as PRD [1] and ChatEval [2])?\n\n[1] PRD: Peer Rank and Discussion Improve Large Language Model based Evaluations https://arxiv.org/abs/2307.02762\n\n[2] ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate https://arxiv.org/abs/2308.07201\n\n[3] AlignBench: Benchmarking Chinese Alignment of Large Language Models https://aclanthology.org/2024.acl-long.624/\n\n[4] Generative Judge for Evaluating Alignment https://arxiv.org/abs/2310.05470\n\n[5] LLMEval: A Preliminary Study on How to Evaluate Large Language Models https://ojs.aaai.org/index.php/AAAI/article/view/29934" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. [results] Did the authors analyze the different types of arguments and justifications between the different ensembles in scoring answers?\n2. [results] Were there any question-answer pairs for which the ensemble methods performed particularly better than the single-judge baseline?\n3. [experiments] How many tokens were needed on average for the different ensembles and models studied?\n4. Section 3.5 is entirely in the appendix, yet referred to in the conclusion [line 499] as discussed. At the minimum, discuss the main results of a section in the main text when referring to it in the conclusion.\n5. [line 504-505] “we have conducted … our framework”: where?\n6. [C.2-C.3] How were any of these chosen? They seem completely arbitrary and unmotivated." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "[clarity] The work was generally easy to read with crisp writing. \n \n[significance] Exploring ways to improve the evaluation of LLM outputs is an important research direction that was well motivated by the authors." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This work explores different processes of rating large language model (LLM) outputs using LLMs. Inspired by legal, psychological, and decision theory, the authors propose two such processes: (1) “Multi-Advocate One-Round Evaluations” (MORE) and “Single Advocate Multi-Round Evaluation” (“SAMRE”). Given a question and two (LLM) outputs, each process uses LLMs in different roles, e.g., as advocates, jurors, or judges, to (iteratively) select the “best” output. The authors further present two theorems respectively claiming that (1) aggregated multi-advocate arguments lead to greater score differentiation than those obtained using iterative debates, and (2) that multi-advocate argumentation requires fewer rounds of interaction to receive the same level of score differentiation as iterative debate. The two processes are tested on the MT-Bench dataset and compared to a baseline of a single LLM judge process using six different LLMs. The authors conclude that their experimental results provide strong empirical evidence for their proposed methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "[originality] Several works have proposed different ways of using LLM ensembles to evaluate LLM outputs. While the authors spend considerable time discussing connections to various disciplines, e.g., decision theory, legal discourse, and psychology, few tangible insights are presented as to how this specific ensemble utilizes these disciplines.\n \n[quality] The experimental results presented in this work simply do not pass the bar for this conference: (1) Only a single, limited dataset is used, (2) critical experimental details are missing, e.g., number of samples used, confidence intervals, temperatures, single-judge baseline setting, length of argument outputs, etc., (3) none of the presented theorems are tested in the experiments, e.g., claims like “greater score differentiation” and “complexity” are neither quantitatively nor qualitatively discussed in the experiments, (4) prompt sensitivity and selection is not discussed at all. This is especially damning for a work focused on improving evaluation.\n \n[significance] In essence, the work proposes (iterative) ensemble scoring using LLMs. The claim of [line 481] “strong empirical evidence for the effectiveness of the proposed LLM advocate architectures in improving the accuracy of LLM output evaluation” is greatly exaggerated and unsupported. There is good reason to believe that most of the reported improvements over a single LLM-as-judge baseline come from the greatly expanded compute budget and the series of hand-crafted prompts. Similar results might thus be obtained by simply providing a single LLM an expanded compute budget and chain-of-thought style reasoning prompts.\n \n[clarity] While the presented theorems and proofs in the appendix are an admirable attempt at introducing rigor to LLM-ensemble evaluation. Yet, they also display a limited understanding of the many practical considerations in using LLMs and the large existing literature documenting poorly understood LLM behaviors. Sweeping, unmotivated assumptions like those on line 321, or the assumption that LLMs assigned different “persona prompts” logically will obtain more diverse and stronger arguments limit the usefulness of the presented theorems." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "see summary" }, "rating": { "value": 1 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "see summary" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "I suspect that this paper may have been generated by a generative AI (such as ChatGPT). The evidence supporting this suspicion includes:\n\n1. The title of the PDF differs from the title listed on OpenReview.\n2. A significant portion of the literature cited appears to be fabricated. While I have not verified every citation, most of the references listed from 2023 onwards seem likely to be fake.\"\n\n\nFor examples:\n\n[10] Wei-Lin Chiang, Zhuohan Li, Zi Lin, Eric Wong, Zihang Zhang, Andy Zou, Lianmin Zheng, Siyuan Yu, Yi Tian, Yinghai Zhu, et al. Chatbot arena: Benchmarking open large language models in the wild. arXiv preprint arXiv:2306.01670, 2024.\n\n[25] T. Lanctot, A. Charnock, and J. Badger. Evaluating multi-agent systems in language models. In NeurIPS 2023 Workshop on Multi-Agent Systems, 2023.\n\n[26] Y. Li, D. Chen, and T. Brown. Agents as evaluators: The role of multi-agent systems in llm assessment. In Proceedings of the 2024 Conference on Neural Information Processing Systems (NeurIPS), 2024.\n\n[34] S. M. Panickssery, E. Lee, and K. Lee. Llm-based evaluators for language models: Opportunities and challenges. In Proceedings of the 2024 International Conference on Learning Representations (ICLR), 2024." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "see summary" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- In the proposed MORE framework, why employ three advocates for each answer? Are these advocates different in any way? Additionally, why does judge J provide scores s_1 and s_2 for both answers at the same time (Line 245)? Does this introduce additional bias? I assume the distributions of s_1 and s_2 obtained this way differ from the distributions obtained if s_1 and s_2 were assessed separately.\n\n- What prompts are used for jurors with different backgrounds? I also question whether merely assigning an identity through the prompt (e.g., \"A retired professor of ethics\") allows the LLM’s evaluation to truly represent the standards of that demographic. This method’s effectiveness requires further validation.\n\n- Could the authors provide an example for the stopping mechanism (Lines 262-263)?\n\n- Why does Algorithm 2 discuss the case of three jurors, when the authors claim five diverse jurors (Line 253)? The authors need to provide the correct version of Algorithm 2.\n\n- Why does the performance of the SAMRE architecture without juries in Table 1 surpass that of SAMRE?\n\n#### Minor Problems\n- The authors should cite reference papers for the theories mentioned in Lines 036-039.\n\n- The authors should clarify the version of the LLMs reported in Table 1. For example, the version of Qwen." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "The writing is relatively fluent." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose a framework that interprets Large Language Models (LLMs) as advocates within an ensemble of interacting agents, allowing them to defend their answers and reach conclusions through a judge and jury system." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The framework proposed by the authors can be seen as an implementation of a multi-agent approach in the field of LLM-as-judges, with limited novelty and contribution to the community.\n\n- There is a lack of detailed description and justification for the proposed framework, with specific issues highlighted in the Questions section below.\n\n- The authors mentioned probabilistic modeling as one of the key contributions in the abstract (Line 018), but only dedicated a single sentence to this aspect in the main text (Line 395).\n\n- The authors conducted only one experiment, comparing the accuracy of their designed framework with a simple baseline, which is insufficient to support their claims. I suggest that the authors add the following experiments and comparison methods:\n - **Comparison methods:**\n - LLMs specifically trained for evaluation, such as PandaLM or Prometheus model.\n - Multiple LLM evaluators using a majority voting strategy.\n - **Experiments:**\n - A comparison of the API and time costs between the proposed MORE and SAMRE frameworks and the aforementioned comparison methods.\n - A performance comparison of the MORE and SAMRE frameworks under different parameter settings (e.g., the number of advocates).\n - A bias analysis comparing the MORE and SAMRE frameworks with the aforementioned comparison methods to demonstrate the claim of being \"unbiased\" (Line 116) and mitigating the influence of strategic behavior and individual biases (Line 119)." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024adversarial,\ntitle={Adversarial Multi-Agent Evaluation of Large Language Models through Iterative Debate},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=06ZvHHBR0i},\nnote={under review}\n}" }, "abstract": { "value": "We propose a novel framework for evaluating large language model (LLM) outputs using LLMs themselves as interacting agents in an adversarial debate system. Our approach casts LLMs as advocates, judges, and juries within a structured courtroom-inspired setting. Advocate LLMs engage in iterative argumentation to refine and critique responses, while judge and jury LLMs moderate and assess the debate. We introduce a probabilistic model using Beta-Binomial distribution to analyze error reduction dynamics in this iterative process. Comparative studies of ranking versus scoring methods for LLM jurors reveal advantages of fine-grained scoring in capturing nuanced quality assessments. Experiments across diverse language tasks demonstrate our framework's superior performance in agreement with human judgments and provision of interpretable feedback compared to traditional evaluation methods. This work contributes a theoretically grounded, scalable approach to LLM evaluation that addresses limitations of existing techniques and adapts to rapid advancements in language AI technologies." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "LLM Evals", "Adversarial analysis", "Mechanism Design" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/f066ebadc0dde761530628129f3872aef2a0db6b.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Adversarial Multi-Agent Evaluation of Large Language Models through Iterative Debate" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- What is the exact definition of the novel complexity measure introduced in this paper? And for which models is this measure well-dfined. The related conversation about compression and motivation from information theory and Kolmogorov complexity is very nice but it's unclear to me exactly how this measure is defined. Is this the content of Algorithm 2? Does the output of Algorithm 2 define the complexity measure? \n - in line 400, can you clarify which subset of grokking experiments you used. And why you used this subset. \n - in line 358 you state \"..we show that regularizing the spectral entropy leads to grokking..\" Is this an overstatement? How exactly is grokking defined quantitatively?\n - In Figure 3, you compare your regularization technique with weight decay. What is the dependence of the proposed spectral entropy regularization on the regularization weight? What behavior do you notice as you apply more or less spectral regularization? It would be nice to see the effect as the regularizaiton of the spectral entropy gradually increases.\n - Does Figure 4 include multiple seeds? Why are error bars not visible in this plot?\n\ntypos/nits\n - in Figure 2. Why include the \"ours\" distinction when all plots are \"ours\". \n - line 372, \"ideas\" to \"ideal\"" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "Understanding the role of model complexity and how it should be measured is an important question in machine learning. This paper takes a good step in this direction and presents a compelling case for a complexity measure which is defined using the minimum description length and ideas from compression and information theory. The paper contributes to a deeper understanding of this 'grokking' phenomenon, which has gotten significant attention in recent years. \n\nThe paper has good theoretical motivation and makes an interesting connection with the concept of grokking in machine learning. Their intrinsic complexity measure and regularization technique are well-grounded in theoretical concepts from information theory. The authors provide clear explanations and justifications for their design choices.\n\nThe paper is logically structured and well-written and supports their theoretical claims with experiments on synthetic tasks, like modular arithmetic, for decoder transformer models." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduce a new complexity measure for neural networks and claim that this complexity measure can be used to explain 'grokking'. \"Grokking\" in machine learning is this idea that neural networks suddenly transition from memorization to generalization long after overfitting the training data. They show that their complexity measure correlates with this 'grokking' and then show how this complexity measure can be used to define a new regularization method which encourages low-rank representations. This regularizer is defined using the spectral entropy of the network weights." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The complexity measure defined and explored in this paper is positioned as a way to 'explain grokking'. \n\nComparison with other complexity measures. The empirical results in the paper are nice. But it would be good to have a fair comparison of how other complexity measures look when measure in the same scenarios. It's unfair to say that this new complexity measure \"explains\" grokking without uncovering a scenario where this complexity measure is able to capture this behavior where others are not. Otherwise, it's unclear if this is just a correlation relationship with the perceived behavior of 'grokking'. \n\nLacking discussion of the cost for computing this complexity measure. If I understand correctly, the proposed complexity measure involves a Bayesian optimization procedure for finding the optimal compression parameters, which could be computationally expensive. It would be nice to address or (ideally) investigating how difficult this measure is. This would enhance the practicality of the approach.\n\nFrom what I understand, this complexity measure is somewhat dependent on the hyperparameters, in particular the per-layer truncation threshold $\\kappa(\\tau)$. It would be nice ot have a detailed analysis even experimentally of the sensitivity to this threshold.\n\nThis paper has some very nice ideas and is worth exploring but it would be good to have a section on Limitations of their approach with an honest assessment in terms of other complexity measures and the degree to which the results are not just correlational with this 'grokking' behavior. \n\nThe paper is carefully written and has a nice re-cap of the relevant ideas from information theory and compression in ML. However, the main message of the paper was at times hard to find. For example, what is the exact definition of this new complexity? I understand it relies on coarse-graning of the network and compression using bzip2 and I think the size of the compressed network is the proxy for the complexity. Is that the definition? This paper would benefit from more clear exposition in this respect." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "Have you tried applying these ideas to more complex datasets, does it compare favorably vs weight decay techniques ?\n\nBzip is not ideal to compress weights... are there other points of comparisons available?\n\nWhat is the efficiency of your compression method? How long does it take to compress?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "Excellent writing, compelling ideas, nice experiments, convincing thesis, possible follow-ups." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper studies the phenomenon of grokking through the lens of complexity theory and rate distortion theory. It proposes ways to compress model weights: \n-- Via a parameter quantization operation, as a twist on ideas of Hinton and Van Camp\n-- Via a low-rank approximation operation.\nThe idea is compress the models up to certain rate distortion thresholds, quantified by the loss. \nThey find that this compression is substantially more powerful than traditional compression methods (bzip) and argue that this is a better approximation of the Kolmogorov's complexity of the model.\nUsing this metric, the authors perform experiments on arithmetic operations and find that the grokking phase is associated with a drop from the complexity peak. Following this idea, they propose a new regularizer that apparently increases the grokking effect.\n\nOverall, this is a very well-written paper that lays out super interesting ideas and presents a compelling thesis and nice experiments. I am not sold on the idea that this is an explanation of grokking, but the observations and the conclusions are overall very interesting and I think this is a valuable contribution to understanding better what happens with grokking and is quite promising to improve learning performance of models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Is it really an explanation of grokking or more some interesting and attractive observations?\nThe experiments with the regularizer are not many." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. How did you set the learning rates for experiments? Does the performance of entropy regularization vary with different learning rates?\n2. While entropy regularization surely helps in compressing the model, I expect that both the usual L2 regularization and the entropy regularization will achieve perfect test accuracy. Could you think of a scenario where the proposed regularization technique offers a clear performance advantage over L2 regularization?\n3. Will entropy regularization also help in training larger models with more complicated datasets, where they often do not have simple representations as one-dimensional numbers?\n4. Could the computational overhead of low-rank optimization become significant, especially when applied to large models? If so, how could we mitigate them?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The paper is generally clear, and easy to read and interpret.\n- The paper provides nice intuitions on building generalizable neural networks, especially from the model complexity perspective.\n- The paper considers an interesting set of techniques for model compression with minimal performance loss, and tests them with experiments." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduce a measure of neural networks’ complexity, and show that grokking could be explained by the rise and fall of the model’s complexity. The authors also propose methods for compressing neural networks via quantization and spectral entropy-based regularization, and empirically demonstrate their performances with modular arithmetic tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the paper considers several promising ideas for model compression, there are a few limitations:\n- While the complexity explanation of grokking is interesting, it seems to overlap with the circuit efficiency explanation proposed by Varma et al. (2023). Although the authors acknowledge that model complexity is not exactly identical to efficiency or parameter norms, the added insights in this area feel somewhat limited.\n- The proposed model compression methods are quite similar to existing techniques on quantization and low-rank approximations, which raises questions about the novelty of the approach. Spectral entropy-based regularization is an interesting idea, but concerns about potential computational overhead and their applicability in more complex settings remain.\n- Lastly, the applicability of entropy regularization techniques in more complex problems beyond the modular arithmetic task raises some concerns. Additional evidence or analysis demonstrating how this technique can advance the complexity-performance Pareto frontier in more difficult tasks will strengthen the paper." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* What's the key motivation of this paper?\n* Could you elaborate on the comparison with bzip2? What is being compressed, problem setup, compressed file size, etc.?\n* What practical implications does this paper have? I would consider a method practically useful if: (1) it can speed up grokking and/or (2) it can compress real-world datasets better than baselines." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "* The paper is well-written and very readable\n* The paper presents \"new\" theoretical tools to analyze neural networks\n* The analysis is a new angle to understand grokking" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes to study the grokking dynamics via the lens of information theory (minimum description length). In particular, they proposed: (1) a new compression algorithm to compress the neural network; (2) a new regularizer based on spectral entropy. They show that the spectral entropy regularizer outperforms the standard weight decay to the extent that a model with lower complexity is obtained. They claimed a factor of 30-40x improvement of the compression ratio over bzip2, which is impressive (although I can't find the file size data). However, none of the compression methods achieve a non-vacuous bound, since models are vastly over-parametrized." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* This paper deals with too many things simultaneously, which makes me a bit lost. What's *the* motivation of this paper? Otherwise, the paper reads like a collection of ok-ish results but none of them is impressive enough. For example, the idea of grokking as compression has been explored by [Liu et al.], [Humayun et al.] and [Deletang et al.]. The idea of using spectral entropy as a measure is explored in [Liu2 et al.], although it is novel to regularize the network with spectral entropy (which is unfortunately expensive).\n* The papers claim a 30-40x improvement in compression ratio, but I did not find and details or data. \n* Although this is a more theoretical paper than an experimental paper, I am not sure about its practical implications.\n\n**References**\n\n[Liu et al] Grokking as Compression: A Nonlinear Complexity Perspective, arXiv: 2310.05918\n\n[Delétang et al.] Language Modeling Is Compression, ICLR 2024\n\n[Humayun et al] Deep Networks Always Grok and Here is Why, arXiv: 2402.15555\n\n[Liu2 et al] Towards Understanding Grokking: An Effective Theory of Representation Learning, NeurIPS 2022" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- Is it possible to perform the same experiments on more complex but still relatively simple datasets like MNIST or CIFAR10?\n\n - Does the generalization bound of Equation (4) only hold for finite hypothesis spaces? If yes is that a realistic assumption in practical learning settings? Moreover, could you be more precise as to why the choice of Solomonoff prior should lead to tighter bounds than other priors, such as the uniform prior over $\\mathcal{H}$?\n\n - Line 181: Why can the empirical risk be understood as the entropy of the data under the model? Is there a way to formalize this fact?\n\n - Is it possible to obtain a formal statement relating the information capacity (Equation (9)) to generalization?\n\n - To what size and precision do the parameters $\\lambda$ and $\\delta$ (Section 4) refer to in practice?\n\n - How would the training accuracy be affected by the addition of Gaussian noise in practical deep learning settings?\n\n - Can you define more precisely the notations used in Algorithm 2, such as BO.SUGGESTPARAMETERS()? More generally, can you provide more details on the Bayesian optimization procedure?\n\n - Does your regularization technique always lead to lower test accuracy compared to weight decay?\n\n - Figures 3 and 5 are not analyzed in the text, can you add some insights on the result they present?\n\n**Remarks/questions regarding lines 152 - 155 and Equation (4)** \nEven though it is not central to the paper, I have some questions about this part:\nAs I understand it, the bounds in terms of Kolmogorov complexity are obtained by choosing a good prior distribution in the bound of Langford and Seeger. It is not clear to me that such a choice of prior provides the most useful bound. More precisely, let $\\mathcal{H}$ be a finite set of hypothesis and $\\sigma : \\mathcal{H} \\to \\mathcal{H}$ be any bijection of $H$. Then $h \\mapsto 2^{K(\\sigma(h))}$ may be used as a prior instead of the usual Solomonoff prior, hence leading to a generalization bound in terms of $K(\\sigma (h))$. Yet another possibility would be to use the uniform prior over $\\mathcal{H}$. Therefore, choice of prior, and therefore the choice of Kolmogorov complexity as a generalization measure, seems to be arbitrary (please correct me if I am mistaken). Can you provide more insights as to why this leads to the most informative bound? \n\nI would be happy to discuss this further, please correct me if I misunderstood something.\n\n\n**Other minor remarks and typos**\n\n - In the introduction, the terms capacity and complexity are used before being defined, which may render the introduction hard to read. In general, more formal definitiosn of these concepts might enhance the readability of the paper. It could also help to define the notion of distortion function.\n\n - Line 122: regulariztion $\\to$ regularization\n\n - Equation (4): there is a missing parenthesis in $\\log(1/\\delta)$\n\n - There might be a clash of notation between the parameter $\\delta$ in Equations (4), (9) and (10). It would be clearer to use a different letter in each of these equations." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- Grokking is an important topic for the community\n\n- The experiments suggest that the proposed regularization technique based on spectral entropy may induce grokking, which may be of practical interest.\n\n- The experiments suggest that the rise and fall of the proposed complexity seems to be predictive of when the model starts to generalize.\n\n- The proposed regularization techniques lead to better generalization bounds than classical weight decay or no regularization." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper studies the grokking phenomenon through compression-based approaches. Inspired by recent work on the intrinsic complexity of neural networks, and combining it with ideas from rate-distortion, quantization and low-rank approximation, the authors propose a new measure of neural networks complexity, which consists essentially of a coarse-graining procedure. They conduct experiments on simple arithmetic tasks which demonstrate that the rise and fall of this complexity might be predictive of the network starting to generalize. Moreover, this leads them to propose a new regularization scheme, based on spectral entropy, whose effect seems to reduce the total description length and the generalization bound, compared to other methods. This might lead to non-vacuous generalization bounds." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Several notions are mentioned repeatedly but without being formally defined, such as capacity, distortion or $(\\lambda,\\delta)$ (Equation (9)). It would improve the paper to include additional theoretical background and more formal definitions. \n\n - It should be made clearer how the quantities introduced in Sections 3.1 and 4 are related to generalization. For instance, is it possible to write down a theorem with explicit dependence on these quantities, or are their consideration partially based on intuitions? Can the link of these quantities with Kolmogorov complexity be made more formal?\n\n - Despite the lack of formal theorems and proofs, the experiments are done on very simple arithmetic tasks. Therefore, it is not clear (neither theoretically nor empirically) whether the results may be generalized to more complex settings. I think that at least one experiment on a small dataset like MNIST or CIFAR10 could improve the paper.\n\n - It would be useful to include an experiment comparing the performance (in terms of accuracy) with and without the proposed regularization scheme. Indeed, we see that it reduces the MDL and the generalization bound, but, if I am correct, it is not clear whether it achieves better performance overall.\n\n - We see in Figure 4 that the proposed regularization scheme achieves the lowest complexity. However, the complexity is computed by Algorithm 2 and the proposed regularization is precisely penalizing the quantity computed by algorithm 2. Therefore it does not seem surprising that it is the lowest. As an ablation study, it would be interesting to make the comparison using other complexity notions. For instance, using the actual test accuracy would be very informative, to see whether the proposed regularization leads to better performance." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We track the complexity dynamics of neural networks during training to understand grokking, using insights from the theory of Kolmogorov complexity." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024the,\ntitle={The Complexity Dynamics of Grokking},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=07N9jCfIE4},\nnote={under review}\n}" }, "abstract": { "value": "We investigate the phenomenon of generalization through the lens of compression. In particular, we study the complexity dynamics of neural networks to explain \\emph{grokking}, where networks suddenly transition from memorizing to generalizing solutions long after over-fitting the training data. To this end we introduce a new measure of intrinsic complexity for neural networks based on the theory of Kolmogorov complexity. Tracking this metric throughout network training, we find a consistent pattern in training dynamics, consisting of a rise and fall in complexity. We demonstrate that this corresponds to memorization followed by generalization. Based on insights from rate--distortion theory and the minimum description length principle, we lay out a principled approach to lossy compression of neural networks, and connect our complexity measure to explicit generalization bounds. Based on a careful analysis of information capacity in neural networks, we propose a new regularization method which encourages networks towards low-rank representations by penalizing their spectral entropy, and find that our regularizer outperforms baselines in total compression of the dataset." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Compression", "Complexity", "Generalization", "Grokking", "Minimum Description Length" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/40834cdeb5c6c11ccfa1eb404acfe0410ee2fa7b.pdf" }, "presentation": null, "primary_area": { "value": "learning theory" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "The Complexity Dynamics of Grokking" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "The authors are encouraged to address the weaknesses identified in this review and provide clarifications on the points raised in the weaknesses section." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The proposal of a deterministic denoising schedule using an SDF is an interesting alternative to traditional diffusion methods. This approach has the potential to enhance temporal consistency and guide the denoising process more directly towards the ground truth action. The ablation studies presented provide evidence supporting the effectiveness of individual components of the proposed method in specific scenarios." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a novel approach to denoising in diffusion models for robot manipulation tasks. The authors suggest replacing the standard noising/denoising process with a Langevin dynamics denoising field based on a signed distance function (SDF). This field serves as a deterministic gradient for denoising, aiming to improve temporal consistency and convergence to the ground truth action. Additionally, the authors introduce an alternative radial loss function to optimize the denoising network. The method is evaluated on RLBench in simulation.\n\nThe paper presents a potentially novel idea by introducing an SDF-based denoising field for diffusion in robot manipulation tasks. However, the clarity of the writing and the consistency of the mathematical formulations require improvement to ensure a thorough understanding of the proposed method. The limited number of trials and the ambiguities surrounding the evaluation metrics raise concerns about the robustness of the results. The authors should address the discrepancies observed in the convergence behavior and provide a more thorough explanation of the method's capabilities and limitations. Addressing the weaknesses and questions identified in this review would significantly strengthen the paper's contribution and impact." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The applicability of the proposed method appears limited to 2D robotics tasks with end-effector movements, such as tabletop manipulation. The authors do not demonstrate how this approach can be extended to other types of actuations, such as gripper control.\n2. The paper seems to present a potential misunderstanding regarding the capabilities of diffusion models. It is suggested that diffusion models may produce the same noisy action for different successful actions. However, diffusion models are capable of learning multimodal action distributions through the denoising process, even in cases of overlapping Gaussians.\n3. Figure 4 raises questions about the convergence behavior of both the proposed method and the standard diffusion model. In scenarios with multiple successful actions (represented by four red triangles), both methods appear to collapse to a single ground truth action. This behavior contradicts the expectation that these models should be able to learn a multimodal distribution and converge to all valid solutions.\n4. The paper lacks clarity on why the proposed method (CIDM) converges to only one ground truth action in Figure 4, despite demonstrating the ability to learn a bimodal distribution in Figure 2. It remains unclear why the method does not capture the four-modal distribution evident in the task.\n\nThere are a number of places in the text where the authors could provide clarification.\n\n### Confusing text\n\n1. Line 83: The statement \"robot manipulation prefers to sample initial actions over the entire action space\" is unclear. It is possible the authors intend to convey that the training data covers the entire action space, but the phrasing is ambiguous and requires clarification.\n2. Line 175: The phrase \"After eliminating the effects of specific successful action $\\hat{y}$\" is vague. It is unclear what is meant by \"eliminating the effects.\" Specifying the mathematical operation, such as marginalizing out $\\hat{y}$, would improve clarity.\n3. Figure 3 caption, line 337: The caption could improve significantly. It states that the list of 14 tasks tasks are \"highly representative.\" Highly representative of what?\n4. The paper lacks details on the experimental setup, particularly regarding the number of trials and seeds used for evaluation. The authors state that results are based on four trials per task, but it is unclear how many random seeds were used to ensure the reliability of the results.\n5. The metric \"success probability\" requires further explanation. If it is calculated based on four trials per task, the possible values should be limited to [0, 25, 50, 75, 100]%. However, Table 2 presents values such as 82.7%, suggesting a different calculation method or a larger number of trials.\n6. Equations 12 and 13 contain an error. The 2-norm $\\|y - \\hat{y}\\|$ cannot be less than a negative number ($c<0$).\n7. Equations 12, 13, and 14 define the denoising field in a way that seems counterintuitive. The denoising field should be $\\epsilon_x(y) = \\hat{y} - y$ to ensure that a single denoising step, $y + \\epsilon_x(y)$, results in the ground truth action $\\hat{y}$. The gradient should point towards the ground truth, not away from it.\n8. Line 285: The authors claim to be learning a denoising field independent of $\\hat{y}$. However, the training data includes $\\hat{y}$, suggesting that the model likely learns $\\hat{y}$ implicitly. This statement requires clarification or justification.\n9. Table 1 caption: The caption states that underlined text indicates \"suboptimal performance for each column,\". Does this mean the second-best performance or some other criterion? Additionally, not every column has an underlined number.\n\nThe paper could benefit from additional explanations and clarifications to enhance the reader's understanding of the proposed method. \n<!-- The authors could have utilized the extra to address some of the ambiguities and provide more detailed insights. -->\n\n## Minor Typos\n\n1. Line 15: \"CIDM\" is used before its introduction in line 144.\n2. Line 77: \"noises supervision signals\" should be \"noise supervision signals.\"\n3. Line 93: \"Additionally, We\" should be \"Additionally, we.\"\n4. Equation 1, line 161: If referring to the DDPM scheduler, the term inside the square root should be $(1 - \\bar{\\alpha_t})$, not $(1 - \\bar{\\alpha_t^2})$.\n5. Line 28: The statement \"Robot manipulation mainly involves two steps, acquiring effective scene representation and predicting correct actions\" oversimplifies the complexity of robot manipulation, which also involves elements of execution on hardware and reactive control." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Can the author provide more detailed explaination of figure1, I'm not sure I understood (b) correctly, especially the blue circles in it.\n\n- Equation 12 and 13 seem to have some typos, I think it should be $\\exists c>0$.\n\n- The value of denoised field (eq 14) is based on the value of 2-norm between noisy action and successful action. The implicit assumption here is that the 2-norm in the action space is well defined. This assumption is not obvious as the common action space may contains position, angle, linear velocity, angular velocity, torque... The 2-norm between two actions doesn't necessarily make sense." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The paper introduces a novel approach to robot manipulation using a diffusion model, addressing limitations of traditional methods by incorporating a consistent denoising field and a radial loss function.\n- Empirical rigor is demonstrated through extensive experiments on the RLBench benchmark, showing clear performance gains over baseline methods. The ablation studies further validate the contribution of each CIDM component, enhancing confidence in the results.\n- By addressing practical challenges in action prediction for complex robot tasks, CIDM enhances the applicability of diffusion models." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents a novel Consistent Iterative Denoising Model (CIDM) aimed at improving action prediction in robot manipulation tasks by addressing issues with diffusion models, specifically noise inconsistency and timestep variations. CIDM introduces two core innovations: (1) a consistent denoising field, which ensures clear denoising directions and temporal consistency across actions, and (2) a radial loss function that emphasizes actions with minimal noise to achieve more accurate iterative denoising." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- While the paper presents a novel application of iterative denoising to robot manipulation, it lacks a theoretical analysis(Like some other articles on diffusion dynamics$^{[1]}$). Highlighting unique theoretical insights or algorithmic innovations would better justify CIDM’s position in the field.\n\n- The introduction of a radial loss function, while conceptually sound, lacks comprehensive theoretical grounding or references to similar existing loss functions used in other domains. This gap makes it challenging to assess the robustness and scalability of the loss. Providing a more detailed theoretical analysis or justifying it with additional related work on spatial consistency in generative models could clarify its effectiveness.\n\n- The current evaluation focuses on RLBench, but it would significantly benefit from testing in other robotic benchmarks or real-world scenarios to assess generalization capabilities. Evaluating CIDM's performance across tasks with varying levels of action complexity, such as multi-step manipulation in dynamic environments, would enhance the robustness claims.\n\n- Temporal consistency is claimed to improve denoising stability across timesteps, but the scalability of this approach remains uncertain for long-duration tasks. Additional evaluations on tasks requiring extended sequences of actions (beyond 100 timesteps) could illustrate CIDM’s scalability and stability in prolonged scenarios.\n\n[1] Liu, X., Gong, C., & Liu, Q. (2022). Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. the paper mentions they use CLIP encoder to extract the embedding from text instructions and image observations. However, it doesn’t mention how they process the robot state information in their framework. Moreover, if there are multiple views, how do they fuse the embedding from different views? The paper needs to add some clarification for those details. \n2. For qualitative results, the author only shows the stack blocks tasks. It would be interesting to see more qualitative rollouts of other tasks. The paper mentions the method is good for the tasks that has multiple success actions. However, the failure case it show when compared to 3D Diffusor Actor in Appendix A.4 is not multi-modal actions. To solidify the paper’s claim, it is better to include some multi-modal actions example and visualize the denoising process." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The paper clearly illustrates the problem, their motivation to propose the new components to diffusion models and the contributions. \n2. The paper provides theoretical analysis to formalize the problem. \n3. The paper shows good results on RL Bench and does ablation studies over the different proposed components to show the importance of each part." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper aims to solve the inconsistent noise supervision issue in diffusion models. The inconsistency comes from two sources. One source is the multi-modal action labels. The other is time-varying noise in denoising steps. They propose a novel consistent iterative denoising model and a new radial loss to address this issue. The proposed method is tested on RL Bench against other baselines." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. In related work, authors list a lot of recent related work in diffusion models. However, some related work is summarized not very clearly. For example, “Inversion by Direct Iteration (Delbracio & Milanfar, 2023) pursues a simpler form to get rid of the limitations of traditional diffusion.” this sentence is confusing because it is not clear to me what things the paper tries to simplify and what limitations they are getting rid of. Another issue is that the paper mentions that the recent work in diffusion models try to speed up the denoising process and provide in-depth analysis of diffusion models. However, these are not directly related to the inconsistency problem this paper tries to solve. Therefore, I think the paper should reorganize this section so that the connection and difference between related work and the proposed method is more clear. \n2. The main advantage of the proposed method as mentioned by the paper is consistent supervision from multiple successful actions (i.e., multi-modality). However, RL Bench demonstrations is not a demonstration dataset that has obvious multi-modality. A recent paper has proposed a dataset benchmark[1] for evaluation of multi-modal behaviors. It would be interesting to see how the proposed methods and the baselines behave in this benchmark\n3. The proposed method’s improvement over previous methods on Multi-view is not very significant with 82.3% average success rate compared to RVT2’s 81.4%. For each task, the proposed method has the highest success rate only in 7 out of 16 tasks. Therefore, it seems that the performance improvement is limited. \n4. In the results sections, the paper only includes the mean but it is reasonable to also include the std for the success rate as this is usually reported in the other papers.\n\n[1] Xiaogang Jia, Denis Blessing, Xinkai Jiang, Moritz Reuss, Atalay Donat, Rudolf Lioutikov, and Gerhard Neumann. Towards diverse behaviors: A benchmark for imitation learning with human demonstrations. In The Twelfth International Conference on Learning Representations, 2024." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "Which architecture did you use for the denoising network? For a fair comparison, it would be helpful to know how the architectures and the number of parameters are controlled across models." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "Diffusion models for robotic manipulation indeed behave very differently from denoising in the pixel space of conventional diffusion models for image generation. Unlike pixel-space diffusion, where values are confined within a compact [0,1] range, gripper pose space is unbounded. This often causes diffusion models to exhibit underconvergent behavior as illustrated in Figure 4. The proposed method appears to offer some mitigation for this important issue." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors argue that the denoising targets of conventional diffusion models are inconsistent, making them unsuitable for robotic manipulation tasks, as they 1) vanish near local optima and 2) are time-varying. To address this, the authors propose the Consistent Iterative Denoising Model (CIDM), which learns from a time-invariant denoising field combined with a radial loss function. In this proposed denoising field and radial loss function, distant GT actions have less influence than closer ones. The authors compare CIDM's performance against state-of-the-art text-conditioned visual robotic manipulation methods, such as 3D Diffuser Actor and RVT2, in the RLBench settings used in PerAct and GNFactor." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "### **Weakness 1. Lacking Probabilistic Justification**\n\nThe authors argue that the score function of conventional diffusion models being zero at local minima is **biased** and should instead always point toward the nearest target. They state:\n>\n> \"The first problem is that the score function $\\nabla_{y_t} \\log p_{t} (y_t)$ is biased as a denoising field... Since the reasonable denoising field always makes noisy action closer to its target successful\naction...\"\n> \nHowever, this claim is debatable. I would contend that CIDM itself introduces bias while conventional diffusion models are unbiased. For a model to be unbiased, the denoising field should be almost zero near saddle points, as there’s no justification for favoring one specific target. In conventional diffusion models, added noise serves to break such ties. Conversely, CIDM imposes a strong preference toward nearby targets among multiple possible answers, making the output highly sensitive to the initial conditions of the denoising process. This arbitrary choice of denoising field introduces bias in CIDM, unless the distribution of initial points is meticulously selected (as in flow-matching models). Alternatively, one could adopt the Annealed Langevin MCMC viewpoint proposed by Song & Ermon (2019). In this case, however, one should carefully choose the form of noise and denoising target so as to guarantee the learned to model to be unbiased. These considerations are not thoroughly addressed in the paper. Consequently, there's no assurance that the samples $y$, generated by CIDM, follow the actual target policy $y\\sim p_{data}(y|x)$. \n\n### **Weakness 2. Claimed Benefit not Well-supported**\nAs discussed in Weakness 1, CIDM introduces bias. However, as demonstrated by the Cold Diffusion paper, neural networks can still produce reasonable samples across various corruption processes, even if biased. Thus, bias isn’t necessarily detrimental when a meaningful trade-off is achieved. However, for CIDM, the specific benefits of this trade-off remain unclear.\n\nFirstly, it is questionable whether the issue presented in Figure 4 is due to inconsistent training objectives. Rather, it could be due to the inference-time denoising scheduler. For instance, I observe that increasing the number of denoising iterations or lowering the temperature at smaller noise scales often resolves the underconvergence issue shown in Figure 4. Better denoising strategies, such as DDIM, could also be an option.\n\nSecondly, the authors argue that conventional denoising target is difficult to learn, and suggest that CIDM alleviates this issue by using a more consistent target. However, I’m not convinced that inconsistency is the only factor at play here. The primary issue could instead be the precision of the action. Diffusion models often struggle with generating highly precise actions due to their inherently noisy and complicated denoising pipeline. In contrast, models specifically optimized for precision, like RVT2, outperform CIDM and 3D Diffuser Actor in precision tasks such as block stacking as suggested in the experimental result. If the authors argue that inconsistent denoising targets hinder learning, they should provide evidence that biasing the target with a more consistent approach indeed reduces learning variance, i.e., by showing that CIDM demonstrates improved data efficiency or lower performance variance across different seeds.\n\n### **Weakness 3. Insignificant Result**\nThe experimental results are not significant, as only 25 test episodes were conducted per task. For the 18 tasks in the PerAct setting, this amounts to 450 trials. With CIDM achieving an 82.3% success rate, the 90% confidence interval is 0.78991 ≤ p ≤ 0.85134. Thus, a 1% improvement over state-of-the-art methods like RVT2 and 3D Diffuser Actor does not offer substantial evidence of CIDM’s superiority.\n\nEven if the reported performance gain holds, it does not sufficiently justify the bias introduced by CIDM. For example, if an expert policy selects a red block with 90% probability and a yellow block with 10%, we would expect the learned policy to favor red blocks proportionally. This expectation does not hold for CIDM. Every generative model has precision-diversity trade-off, and the RLBench success rate primarily measures precision over diversity. Therefore, sacrificing sample diversity for only a 1% performance gain does not make a lot of sense for me." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024consistent,\ntitle={Consistent Iterative Denoising for Robot Manipulation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=07ZaA3MiL0},\nnote={under review}\n}" }, "abstract": { "value": "Robot manipulation in complex scenarios usually involves multiple successful actions, which requires generative models to estimate the distribution of various successful actions. \nIn recent years, the diffusion model has been widely studied in many robot manipulation tasks.\nHowever, the diffusion model experiences inconsistent noise supervision across various action labels and denoising timesteps, which compromises accurate action prediction.\nOn the one hand, CIDM designs new noise supervision to avoid interference between different successful actions, leading to consistent denoising directions.\nOn the other hand, CIDM unifies all denoising timesteps, avoiding inconsistent predictions of the diffusion model over different timesteps.\nMoreover, we also designed a novel radial loss to make the model focus on denoising results rather than iterative process routes.\nOur method achieves a new state-of-the-art performance on RLBench with the highest success rate of 82.3\\% on a multi-view setup and 83.9\\% on a single-view setup." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "robot manipulation", "consistent iterative denoising", "diffusion model", "imitation learning" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/444137bba6fc98bf0f71907783db1f09bf5eeaef.pdf" }, "presentation": null, "primary_area": { "value": "applications to robotics, autonomy, planning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Consistent Iterative Denoising for Robot Manipulation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "(Q1) L154: Do you also include model textual response \\hat{w}_i, i = {1,..,H} in w_i?\n\n(Q2) Eq 1: \\hat{w} is overloaded as both the model textual response and the embeddings of text inputs\n\n(Q3) Eq 1: Each of w_i might have a different number of tokens (different from n). Do you pad them to n or is the index (n_i) dropped for brevity? That is: (w_i^1, w_i^2….w_i^{n_i}) instead of just (w_i^1, w_i^2….w_i^{n})\n\n(Q4) L222-L225: The observed phenomenon is not clear here. Referring to the appendix also doesn’t add more details, apart from the empirical observation. Can the authors describe this with an example? \n\n(Q5) VLA4CD (no-language): What is the architecture, inputs for this model? Ideally, the human question in the input and the {s_t^{l+1}, .., s_t^{l+n}} must be removed while training.\n\n(Q6) L413: How did you balance the two losses for DriverGPT4? Did you have a hyperparameter search for the loss weights similar to your approach?\n\n(Q7) L477: The reasoning here is heavily dependent on the discretization strategy used for each environment. How were the actions discretized for this environment? Was there a hyperparameter search performed to get the best strategy?\n\n(Q8) L140-142: How is this problem avoided in the current setup? It’s not clear in the text here.\n* Action space dimension is small, i.e., 2 (acceleration and steering) How does this scale with more variables?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "(S1) In general, the intended direction of this work, i.e., a model that can take actions while retaining the ability to generate textual responses to a user is useful. Please see weaknesses for further discussion.\n\n(S2) The technical details as presented in the paper are easy to understand and follow." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The current manuscript proposes to build a large language model (LLM) capable of understanding multiple modalities like text, vision, and actions; and producing them as outputs. In particular, it develops a Visual Language Action model for Chatting and Decision Making (VLA4CD) that produces continuous actions without losing its ability to chat with a user simultaneously. Notably, the action space is not discretized and kept continuous, unlike prior works in this area. The paper also demonstrates experiments on CARLA dataset to claim that this approach is effective and can provide real-time decision making compared to prior art, while retaining its text interaction capability." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "(W1) The current manuscript suffers from a clear lack of motivation for why we need a model that can produce both actions and also “chat” (L21, for instance) with a user. There are two main problems here:\n* Throughout, the ability of “chatting with people” (L88) has not been characterized well. It is not open-ended dialog on any topic but rather an explanation of what actions to take or why it has taken a certain action in a given situation. This is misleading as currently phrased.\n* Much of the motivation is around “a human driver can operate while chatting with a friend”, which does not apply to why we need a unified model. For instance, why not have an actuation model and an open-ended dialog model in the autonomous vehicle to achieve the above desired motivation? This indicates the lack of a clear motivation from an application standpoint.\n\n\n(W2) Even if one were to scope the “chatting with users” ability down to producing explanations as responses to a fixed set of templated questions (see A.10), the manuscript does not follow through via corresponding experiments. Both actions and text-generation capability has been evaluated independently, once again begging the question as to why such a unified system is useful. There are no experiments to verify the following:\n* The model actually actuates based on the textual outputs? I.e., if the model responds with “I will take the right lane in 20 mins”, does it actually do that?\n* Are these textual explanations correct/sound given the state of the environment?\n* What is the correlation of the GPT-4o score evaluation with human evaluation? \n\n\n(W3) There are some concerns around the experimental validation of the proposed methodology:\n* The reported experiments on town03 and town04 from the CARLA environment do not seem to match with any of the existing benchmarks with prior works (C, D). \n* To further exacerbate this issue, none of the baseline results are from literature and have been reported based on reproductions in this work. \n* Missing baselines, see [A] for more information.\nThis raises serious questions about the efficacy and usefulness of the proposed methods from an empirical standpoint. Why were existing, standardized benchmarks not used for model comparisons? Request the authors to address these concerns without which the benefits of this approach will remain unclear.\n\n\nReferences\n* [A] DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Driving. https://arxiv.org/pdf/2312.09245.\n* [B] Think2Drive: Efficient Reinforcement Learning by Thinking with Latent World Model for Autonomous Driving (in CARLA-v2). https://arxiv.org/pdf/2402.16720\n* [C] CARLA Autonomous Driving Leaderboard. https://leaderboard.carla.org/\n* [D] TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving. https://arxiv.org/pdf/2205.15997" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "- How does the model compare to DriveGPT4 and why does DriveGPT4 do so bad? DriveGPT4 is doing exactly same thing as this model aims to do (text generation + action generation).\n- Why are there no use case of the model actually chatting? How do the authors define chatting? The example of the introduction mentions the authors are inspired by human driver talking to a friend while driving, but the model doesn't actually engage in free form chat that goes beyond a single step.\n- How does author plan to make the model robust to noise when exposed to unrestricted chat?\n- Why should we add language generation capability to VLAs? The motivation for that seems non-existent in this paper and there's no novel use case of the generated language.\n- Why do the authors think using separate loss for language and action generation (unlike DriveGPT4) improves the decision making performance?\n- Why do the authors only focus on the self-driving task?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The main contribution of this work is that the authors add a language generation capability to VLA in self-driving scenarios.\n- The paper is well-structured, making it easy to read and understand the authors' approach.\n- The finding that separating language and action prediction loss can improve decision-making is a significant contribution that provides valuable insights into how VLAs can be effectively trained. It is encouraging to see that this is empirically demonstrated to be useful in self-driving scenarios. However, it is concerning that introducing some language noise into the training dataset can have a considerable impact on decision-making processes. Since real-world datasets will inevitably contain substantial noise, developing methods to ensure robustness against such noise is essential for the model's practical application." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "- The authors propose VLA4CD, a model for self-driving that is capable of generating both language and action tokens. \n- VLA4CD is trained using three different types of loss functions: language generation, action generation, and image reconstruction. \n- The trained model demonstrates superior performance in both decision-making and question answering compared to models such as DriverGPT4 and OpenVLA in the gym-carla environment." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- I don't quite understand why VLAs need to chat based on the author's motivation. Chatting is an inherently multi-turn conversation with a specific topic, but example of such capability of the model is completely lacking. I wonder what the authors' definition of chatting is. The model doesn't actually \"chat\" but simply outputs action description. It is far from the example in the introduction where authors want to build a model that can talk with a friend while driving.\n- Text generation has already been explored with DriveGPT4. In this paper, text generation is not used for any novel applications other than simply translating action tokens into language. I fail to understand why does the author claim text+action generation is something novel since there's already a model that does it. \n- Adding chat capabilities could potentially make the model less robust when exposed to noise in language interactions. Since the model learns to associate language with specific action tokens, any slight disruption to that association (e.g., due to noise) could significantly impair its action prediction performance. If the model can engage in unrestricted conversation, it is likely to encounter more noise, which could seriously affect its decision-making abilities, which is the most important goal of VLAs. It might be more effective for VLAs to focus solely on action prediction and incorporate chat functionality with separate models. With the current motivation, it seems there is no strong rationale or necessity to integrate chat capabilities into VLAs.\n- Following the point above, I feel like the paper could be better framed on how to manage loss when training VLAs, which is a much interesting topic.\n\nTypo: Figure 1 interatcion" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. What is the average number of images per training and inference case?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "1. The problem that developing a chatting and simultanenous decision making, itself is underexplored and important.\n2. The proposed model gives both reasonable question answering output and reasonable action output." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper opposed to simulating MLLMs as human in a real-world situation that require both chatting and decision making. For example, human drive can drive safely while having conversations with passengers. This is an important application problem in autonomous driving system." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. I don't buy the idea that by simply concatenating the question answer data and action prediction data together and supervised finetune the LLaVA-like MLLM can solve the proposed problem. As described in the abstract, driving and chatting simultaneously, however the proposed LLM-based model is autoregressive decoding. From the architecture overview, we can see this proposed model can only provide a prediction after a very long answer output. No inference speedup or simultaneous decoding technology is being used or proposed to achieve this.\n2. The only contribution to me seems combining action prediction and question answer data, which is very trivial. No siginificant improvement is achieved compared with specialist model in each single task. And the approach to combine these two tasks chatting and decision making tasks are not actually achieving prediction both but sequentially.\n3. The paper neeeds to be revised in the writing. Such as notions in Figure 1 is totally missing. Training and architecture details are missing in Experiments section, etc." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose an approach to build a pre-trained multimodal model for simultaneously chatting and decision-making." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024how,\ntitle={How to Build a Pre-trained Multimodal model for Simultaneously Chatting and Decision-making?},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=07cehZ97Xb},\nnote={under review}\n}" }, "abstract": { "value": "Existing large pre-trained models typically map text input to text output in an end-to-end manner, such as ChatGPT, or map a segment of text input to a hierarchy of action decisions, such as OpenVLA. However, humans can simultaneously generate text and actions when receiving specific input signals. For example, a driver can make precise driving decisions while conversing with a friend in the passenger seat. Motivated by this observation, we consider the following question in this work: is it possible to construct a pre-trained model that can provide both language interaction and precise decision-making capabilities in dynamic open scenarios. We provide a definitive answer to this question by developing a new model architecture termed Visual Language Action model for Chatting and Decision Making (VLA4CD), and further demonstrating its performance in challenging automonous driving tasks. We build VLA4CD on the basis of transformer-based LLM architecture. Specifically, we leverage LoRA to fine-tune a pre-trained LLM with data of multiple modalities covering language, visual, and action. Unlike the existing LoRA operations used for LLM fine-tuning, we have designed new computational modules and training cost functions for VLA4CD. These designs enable VLA4CD to provide continuous-valued action decisions while outputting text responses. In contrast, existing LLMs can only output text responses, and current VLA models can only output action decisions. Moreover, these VLA models handle action data by discretizing and then tokenizing the discretized actions, a method unsuitable for complex decision-making tasks involving high-dimensional continuous-valued action vectors, such as autonomous driving. The extensive experimental results on the closed-loop autonomous driving platform CARLA validate that: (1) the model construction method we proposed is effective; (2) compared to the state-of-the-art VLA model, VLA4CD can provide more accurate real-time decision-making while retaining the text interaction capability inherent to LLMs." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "vision language action model; decision making; autonomous driving; multimodal" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/55d50e8073e31524d6e9cef3c6f6f2b6202a851a.pdf" }, "presentation": null, "primary_area": { "value": "applications to robotics, autonomy, planning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "How to Build a Pre-trained Multimodal model for Simultaneously Chatting and Decision-making?" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "It says “We generate data for pairs D_{Ei, Ej} and triplets D_{Ei, Ej, Ek} in our experiments”. I wonder if the authors have any intuition about how performance changes with the size of subset k." }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "* The problem the work addresses is important.\n* Experimental results show that this method scales better than simple paraphrasing or direct pretraining, and that retrieval-augmented generation further boosts performance of this model. \n* The authors also present a theoretical model explaining EntiGraph’s log-linear scaling pattern, providing insights into the mechanics of synthetic data’s impact on learning efficiency.\n* Paper is clear and well-written." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper addresses the problem of data inefficiency in pretraining language models. Current pretraining corpora may not generalize effectively and models may benefit from structured, repeated, diverse representations of knowledge. \n\nThe proposed is a two-step process that (1) extracts entities from the corpus and then (2) extracts relationship information amongst a subset of the entities.\n\nExperimentation uses the QuALITY corpus and dataset, which is a benchmark for long-document reading comprehension. Evaluation compares with relevant baselines like training on the original QuALITY corpus and a corpus containing rephrasings." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the experiments focus on the QuALITY corpus, it remains unclear how well this would apply to other domain-specific corpora or more complex fields (e.g., legal or math data)." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "For the data generator, what type of models are necessary to have good performance? (why use GPT4 and not open-source models)\nThe paper shows that the generated data is useful, but how does it look like? (is it good quality text, factual, natural looking, ...) \nWhat is the significance of section 6?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "The paper does a good job at demonstrating the benefit of the synthetically generated data, by including relevant natural baselines. \nThe proposed method seem to work well and can be useful for continued pre-training tasks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a method to continue pretraining LLM with a synthetic data augmentation method. The method is based on expanding the training corpus with many verbalizations of the entity graph present in the training corpus. It moves from a sparsely verbalized entity graph to a more densily verbalized one by using only the source documents and prompting LLMs to generate the new tokens.\n\nThe paper shows that the method is beneficial for downstream tasks in closed- and open-book QA as well as RAG. \n\nOverall, I think the paper is worthy of acceptance, it propose a clean method with good results and the experiments are fairly convincing." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The work relies on commercial and closed-source models (GPT4) for generating the synthetic data, making this work non-reproducible. Since the data generation process is the central contribution, it would have been interesting to have insights about how well different models can perform this data generation task. \n\nThe paper proposes only extrinsic evaluation of the generated data but does not provide intrinsic measures, i.e., how good is the generated text? \n\nIn my opinion, section 6 is not particularly useful. It is unnecessarily mathematical, based on simplistic assumptions and does not bring useful insights (For many continuously increasing lines, there anyway exists a mixture-of-exponential that fit it)" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. How sensitive is the synthetic pretraining process to the specific hyperparameters used for entity extraction and relation analysis? Would tuning these parameters significantly affect the generated corpus quality?\n\n2. How does the synthetic corpus compare to a manually curated dataset in terms of quality and impact on downstream tasks?\n\n3. Could EntiGraph be used effectively in scenarios where entities are ambiguous or domain-specific (e.g., medical or legal texts)?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The proposed EntiGraph approach for generating synthetic data is well-motivated and demonstrates clear improvements in downstream performance, as shown by the experimental results.\n2. The paper includes comprehensive evaluations, including closed-book QA, instruction following, and open-book settings. The results show a significant performance improvement over baselines, validating the effectiveness of synthetic pretraining.\n3. The authors provide a theoretical analysis of EntiGraph's effectiveness, which aligns well with empirical observations and provides a deeper understanding of its scaling properties." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes \"synthetic continued pretraining\" to enhance language models' domain adaptation efficiency, particularly when adapting to a small corpus of domain-specific documents. The authors introduce EntiGraph, a synthetic data augmentation algorithm that constructs a knowledge graph from the entities in the original corpus and synthesizes diverse text to enhance the pretraining process. The approach was evaluated using a reading comprehension dataset, QuALITY, showing notable improvements in model performance with both closed-book and open-book scenarios." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The evaluation relies on the QuALITY dataset, which may not be representative of all types of small corpora. A broader range of datasets, particularly from diverse domains, would make the results more generalizable.\n2. Although the authors attempt to mitigate hallucinations by grounding synthetic data generation in the original corpus, the risk of generating inaccurate information is inherent in using a language model for synthetic generation. This aspect needs further empirical examination, such as quantitative metrics to evaluate hallucination rates.\n3. The approach relies on using strong language models like GPT-4 for synthetic data generation. The practical feasibility of using this approach might be limited if users do not have access to such models due to their computational cost. What if it was replaced with LLama 3 8B?\n4. While the paper includes useful baselines such as \"Rephrase CPT,\" more comparisons with alternative data augmentation or synthetic generation methods from recent literature could strengthen the claim that EntiGraph is an effective strategy." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "The paper could be more robust if you had more than just the QuALITY dataset. It is a perennial problem to find hard datasets to work with, so I understand this may be all there is for now, but given the chance I would attempt to reproduce the results on a different set. The authors mention linear algebra (a much harder topic, I think): is there any corpus for that subject?\n\nThe presentation of how exactly you generate the text to train Llama 3 8B with EntiGraph is still a little fuzzy to me, in particular it would be nice to see some examples of what you generated. It is helpful to have the prompts, but some output always grounds the presentation. \n\nFinally, I imagine GPT-4t made errors in producing the training data--did you search for these? Even at a quick glance how often did it make errors, and what, if anything, did you do to filter them out?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The experiments are convincing that the EntiGraph approach improves the LLM's ability to accurately answer questions about a small corpus. In particular the closed-book results in Figure 3 show that the EntiGraph approach leads to far more salient claims per false claim than any of the other models, including GPT-4, or training the LLM (Llama 3 8B). The benefit is substantially less in the open-book RAG case, but there is still substantial improvement. The theoretical model to explain how the model improves QA accuracy with increasing tokens provides some good intuition as to how the model learns. \n\nOverall the text is clear and easy to read." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper addresses how to train LLMs in a data scarce regime, given that LLMs require O(1000) examples of a fact to actually \"learn\" it. This has applications both to niche corpora (e.g., mathematics textbooks) as well as to training larger models once all human text is exhausted. The authors propose to use a pre-trained LLM to (1) extract entities and summaries from a comparatively small, niche corpus, and (2) use the extracted entities to generate rephrased assertions about those entities, to facilitate learning by a second (here, smaller) LLM. They experiment with a 1.3M token reading comprehension dataset, and test the approach against several baselines, including closed-book tests on the LLM used to extract entities and the rephrased text used to train the second LLM. Finally, the authors present a mathematical model through which they attempt to understand the training behavior of this data augmentation system." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I still have reservations that there is some amount of distillation of GPT-4 into their Llama 3 8B: it seems possible to me that a RAG-prompted GPT4 could generate additional information that is somehow \"unlocked\" by the RAG prompt, but which the closed-book version was unable to access. At the risk of anthropomorphizing, this is akin to a human getting a visual or audio cue and suddenly recalling whole complex memories. It would make the paper stronger to dig into the results of entity extraction and the generated text to see whether it is rephrasing/paraphrasing, or whether possibly actual new information is injected.\n\nEven so, it would have helped this reader to have pointed out the significance of the closed book experiments earlier on. It isn't stated explicitly until the Limitations section.\n\nI don't feel particularly qualified to check your proofs of theorems, and moreover I think the main value of the theoretical model is to help the reader understand intuitively why the approach works (these may be connected observations). Is all of the theory necessary? Perhaps a simulation would do as well?\n\nAnother issue is that much of the benefit of the approach vanishes (though not completely) when using a RAG model directly. Is this approach worth the extra training, given the modest gains? The core problem, really, is how many examples LLMs take to learn anything well. This paper finds a way to side-step that successfully, but doesn't solve it directly." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024synthetic,\ntitle={Synthetic continued pretraining},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=07yvxWDSla},\nnote={under review}\n}" }, "abstract": { "value": "Pretraining on large-scale, unstructured internet text enables language models to acquire a significant amount of world knowledge.\nHowever, this knowledge acquisition is data-inefficient---to learn a fact, models must be trained on hundreds to thousands of diverse representations of it.\nThis poses a challenge when adapting a pretrained model to a small corpus of domain-specific documents, where each fact may appear rarely or only once.\nWe propose to bridge this gap with synthetic continued pretraining: using the small domain-specific corpus to synthesize a large corpus more amenable to learning, and then performing continued pretraining on the synthesized corpus.\nWe instantiate this proposal with EntiGraph, a synthetic data augmentation algorithm that extracts salient entities from the source corpus and then generates diverse text by drawing connections between those entities.\nSynthetic continued pretraining with EntiGraph enables a language model to answer questions and follow generic instructions related to the source documents without access to them.\nIf the source documents are instead available at inference time, we show that the knowledge acquired through our approach compounds with retrieval-augmented generation.\nTo better understand these results, we build a simple mathematical model of EntiGraph, and show how synthetic data augmentation can \"rearrange\" knowledge to enable more data-efficient learning." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "large language model", "synthetic data", "continued pretraining" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/3f7f02cad5174828e2ae4812f3a70a8db2a199cd.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Synthetic continued pretraining" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "See weaknesses." }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- The problem is important and well-motivated, as affordance grounding is crucial for robotic manipulation and human-object interaction understanding\n- The proposed pseudo-labeling approach effectively leverages existing foundation models (VLpart, SAM) to provide supervision, addressing limitations of previous CAM-based methods\n- The label refinement process using exocentric images is novel and well-designed, providing a clever way to improve initial pseudo labels\n- The reasoning module helps generalize to unseen objects, which is crucial for practical applications\n- The writing is clear and the method is well-explained with appropriate visualizations" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper tackles weakly supervised affordance grounding (WSAG) by leveraging foundation models to generate pseudo labels, departing from previous CAM-based approaches. The authors propose a three-stage pipeline: (1) using VLpart and SAM to generate initial pseudo labels by mapping affordance-object pairs to part names, (2) refining these labels using human-object interaction cues from exocentric images, and (3) training an affordance grounding model with the refined pseudo labels. The method also includes cross-view feature alignment and a reasoning module to handle unseen objects. The approach shows significant improvements over existing WSAG methods" }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The choice of CLIP as the vision encoder could be better justified given previous work suggesting limitations (vs DINO, OWLViT, SAM). For example, the paper will be stronger with an ablation study of different visual encoders." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Could you provide more details about failure cases and limitations of the proposed approach?\n2. How sensitive is the method to the results of VFM? How well can the refine state correct possible errors by VLpart and SAM?\n3. How does the computational cost (training & inference) compare to existing CAM-based methods?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- Clear writing and organization.\n- Well-motivated technical approach with clear problem formulation.\n- This paper propose a novel approach that uses visual foundation models and part-level semantic priors for WSAG, unleashing the power of these models for affordance learning.\n- Using human occlusion cues for label refinement, which is an innovative insight.\n- Comprehensive experimental validation and thoughtful analysis of limitations in existing methods." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper addresses the task of weakly supervised affordance grounding (WSAG), where the goal is to identify affordance regions on objects using only image-level labels and human-object interaction images. \nThe key contributions include:\n- A novel pseudo-supervised training framework and pipeline that leverages visual foundation models to generate affordance heatmaps, mapping affordance classes to object parts.\n- Three key enhancements to improve performance:\n - Label refinement using interaction cues\n - Fine-grained object feature alignment with exocentric images\n - Reasoning module for better generalization\n- Extensive experiments demonstrating significant performance improvements over existing methods" }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Could benefit from more analysis of failure cases.\n- The label refinement stage using human occlusion cues may be problematic when interactions are ambiguous or when multiple affordances exist.\n- The mapping from affordance to part names is ad-hoc and manually crafted, which limits the scalability to new affordance types and more complex objects." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Aligning the features of an object from different views is a commonly used strategy for feature learning. How is this strategy related to pseudo label generation and refinement.\n2. Some designs need more detailed ablation studies. E.g., how does the proposed fine-grained feature alignment process with SAM perform when compared with the previous work aligning the features directly. Is there any significant performance difference?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. Training affordance grounding models with object labels is an interesting question.\n2. Using off-the-shelf foundation models to generate affordance label is an interesting idea.\n3. Experiments show promising results." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper propose a weakly supervised affordance grounding framework. It uses off-the-shelf foundation models to generate pseudo labels of object parts. To further improve the performance, a label refining strategy, a fine-grained feature alignment process, and a lightweight reasoning module are introduced. Experiments show promising results." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. As shown in the ablation study table 2, the improvements of using all these three modules look marginal over using one module. It seems that the effectiveness of the three components are not significant.\n2. In section 3.4, the authors propose to align the features of exo- and egocentric images after SAM segmentation while the existing methods directly align the features of the two images. However, there is no solid experiments to show the effectiveness of this design.\n3. The framework refines the affordance labels with the need of the corresponding exocentric image which may not be available sometimes." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose a pseudo-label based method for weakly supervised affordance grounding, utilizing the semantic priors of vision foundation models." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024weaklysupervised,\ntitle={Weakly-Supervised Affordance Grounding Guided by Part-Level Semantic Priors},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0823rvTIhs},\nnote={under review}\n}" }, "abstract": { "value": "In this work, we focus on the task of weakly supervised affordance grounding, where a model is trained to identify affordance regions on objects using human-object interaction images and egocentric object images without dense labels. \nPrevious works are mostly built upon class activation maps, which are effective for semantic segmentation but may not be suitable for locating actions and functions. Leveraging recent advanced foundation models, we develop a supervised training pipeline based on pseudo labels. The pseudo labels are generated from an off-the-shelf part segmentation model, guided by a mapping from affordance to part names.\nFurthermore, we introduce three key enhancements to the baseline model: a label refining stage, a fine-grained feature alignment process, and a lightweight reasoning module. These techniques harness the semantic knowledge of static objects embedded in off-the-shelf foundation models to improve affordance learning, effectively bridging the gap between objects and actions.\nExtensive experiments demonstrate that the performance of the proposed model has achieved a breakthrough improvement over existing methods." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "weakly supervised affordance grounding", "foundation model", "pseudo label" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/8e86da678f19c14a55c806dd8360c86e09890140.pdf" }, "presentation": null, "primary_area": { "value": "applications to computer vision, audio, language, and other modalities" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/6d4a8cc6f8e2fdea56ed70c51ffccd79c27af4ac.zip" }, "title": { "value": "Weakly-Supervised Affordance Grounding Guided by Part-Level Semantic Priors" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see the weakness.\n\n1. The key question is that we can use fusion of other existing metric without introducing a new one.\n2. The formulation of clipscore is not common." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The author shows the disadvantage of ClipScore, and try to design a new one. \n\n1. The key question is that we can use fusion of other existing metric without introducing a new one.\n2. The formulation of clipscore is not common." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In this paper, the author proposes a new metric called AugClip, for Text-guided Image Editing." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Some notations are confusing. \nT sometimes is text, while T sometimes is the set. \n\n2. Overclaims in contribution-1 \"We are the first to point out CLIPScore’s reliability in text-guided image editing\"\nIn fact, most researchers recognize this point, and use a cocktail metric, like FID + Clipscore + SSIM + human evaluation. \n[a] Holistic Evaluation of Text-to-Image Models. NeurIPS. \n\n3. Conflicts in contribution-2 \nIn abstract, the author said using MLLM. \nIn introduction, the author claims LLM. \n\n4. Contribution-3 said \"demonstrates\" but there is no mathmatic proof. \n\n5. Figure 1 does not convince me. \nWe could simply use cosine(f_source, f_edit) to see the preservation. \nhttps://openaccess.thecvf.com/content/WACV2024/papers/Tanjim_Discovering_and_Mitigating_Biases_in_CLIP-Based_Image_Editing_WACV_2024_paper.pdf\n\n6. Figure 2 is similar to Figure 1. \nWe could simply use cosine(f_source, f_edit) to see the preservation. \n\n7. Eq.1 is not commonly-used. \nCould you show the reference? It does not make sense, since clip feature can not use plus or minus operation. \nMost cases I read is using cosine(f_modification text, f_editted image) \n\n8. One simple ablation is missing. \nHow about the weighted sum like cosine(f_modification text, f_editted image) + 0.5*cosine(f_source image, f_editted image) ?\ncosine(f_modification text, f_editted image) higher is better modification. \ncosine(f_source image, f_editted image) higher is better preservation. \nUsually, we will use the FID to indicate the preservation as well. \n\n9. I am confusing about Eq.3,4,5. \nEq 4,5 is about a , not v.\nEq 3 is about v, not a. \nBut the author said a can control v. \"the refined version of v is obtained through hyperplane optimization using as and at.\"" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. The main contribution of this paper is the new metric, augclip, for text-guided image editing. However, very few editing methods are tested with this new metric, which weakens the solidness of the metric. Considering the limited rebuttal period, I will not ask the authors to evaluate all editing methods in table 6 in the appendix. However, since the main benchmark illustrated in the paper is TEdBench[2], I suggest the authors to evaluate Forgedit[3] with augclip since they also released the complete editing results of TEdBench on github. You have to compare image editing methods with augclip to demonstrate its effectiveness instead of text-to-image models like stable diffusion itself in your paper. \n\n2. The clip score definition in equation 1 is different from the mainstream reference[1]. Where does this equation 1 come from? Why is it used instead of [1]?\n\n3. How long does it take to train the augclip metric on each benchmark? \n\n\nI am willing to raise my rating score if the authors could stress my concerns in the revised version of this paper. \n\nreferences:\n\n[1] Clipscore: A referencefree evaluation metric for image captioning. In EMNLP\n\n[2] Imagic: Text-based real image editing with diffusion models. In CVPR\n\n[3] Forgedit: Text-guided Image Editing via Learning and Forgetting. In arxiv" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The observation and visualization that clip score favors modifications instead of preservation is interesting.\n\n2. The idea to transfer the trained SVM from text space to image space and compute the minimum v from source image embedding to target prompt embedding is clever. \n\n3. the two-alternative forced choice testing and ground truth testing are reasonable." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposed a novel evaluation metric, augclip, for text-guided image editing. Motivated by the observation that clip score biases towards modification instead of preservation, the authors utilized GPT-4v to rephrase the source prompt and target prompt to extract essential visual attributes in the form of text prompts. Then the authors trained classification models to classify source prompts from target prompts. Since CLIP itself aligns the image space and text space, the classification model trained with source prompts and target promtps could also be utilized to compute the minimum vector v that transfer source image embedding to target text embedding. The augclip metric is then defined to be the cosine similarity of image embedding of the edited image and the vector sum of v and source image embedding. The proposed augclip demonstrate superior alignment with human evalutions than clip score and lpips score." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Since the main contribution of this paper is the augclip evaluation metric, the authors should compare with more than one image editing method on each benchmark. However, the authors only evaluated the results of one image editing method, the results of stable diffusion 1.5 (which is just a text to image base model without editing capability itself) and the original reference image, with the proposed augclip. For a new evaluation metric, this is far from enough. For example, the authors showed the reference images from TEdBench [2] multiple times in the paper, yet they only evaluate the scores of Imagic+Imagen. There are other related works on this benchmark, for example, Forgedit[3] open-sourced their implementation and released their results on TEdBench on github. \n\n2. Incorrect clip score definition. The clip score in this paper, shown in equation 1 in section 3.1, is different from the usual clipscore being used in text-guided image editing literature [1]. For example, in Imagic[2] and Forgedit[3], the clip score metric's definition follows [1]. \n\n3. Most editing methods in table 6 in the appendix never appear in the paper and section 4.3 is not written well thus is very confusing.\n\n\n\n\n[1] Clipscore: A referencefree evaluation metric for image captioning. In EMNLP\n\n[2] Imagic: Text-based real image editing with diffusion models. In CVPR\n\n[3] Forgedit: Text-guided Image Editing via Learning and Forgetting. In arxiv" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Please refer to **Weaknesses**." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The authors evaluate AugCLIP on multiple benchmarks, demonstrating that AugCLIP outperforms CLIPScore and LPIPS with various editing methods and datasets.\n\n- AugCLIP can evaluate both the modification and preservation of the editing images. Compared to CLIPScore, AugCLIP is a more comprehensive metric.\n\n- By leveraging GPT-4V, AugCLIP can evaluate more fine-grained differences between the ground truth image and the edited image. \n\n- The authors provide ablation studies to evaluate difference components of AugCLIP." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces AugCLIP, an evaluation metric designed for text-to-image editing tasks. AugCLIP aims to address limitations in CLIPScore, which can not evaluate the preservation of the original input image. The method leverages GPT-4V for detailed descriptions of the source and target images. By creating a \"modification vector\" based on source and target attributes, AugCLIP balances preservation and modification. The authors demonstrate that AugCLIP outperforms metrics such as LPIPS, CLIPScore on various datasets." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- It seems that the authors overclaimed their contributions. For example, in Line 081-083, they mentioned that \"We are the first to point out CLIPScore’s reliability in text-guided image editing\". As far as I know, many papers have pointed out the limitations of CLIPScore. Almost all image editing methods leverage CLIP to evaluate the modification, and LPIPS/FID to evaluate the preservation. For example, [1,2] provides both CLIP and LPIPS to evaluate the editability–fidelity tradeoff.\n\n- The authors seem to confuse CLIP score with CLIP directional similarity score (*i.e.*, directional CLIP loss). From my understanding, the definition in Section 3.1 is more like CLIP directional similarity score rather than CLIP score. Please double check the definition of CLIP and CLIP similarity score in the following link: \n https://huggingface.co/docs/diffusers/conceptual/evaluation.\n\n- The experiments only involves metrics like LPIPS, CLIP. Please consider include the tradeoff between CLIP and 1-PIPS or FID.\n\n- Introducing GPT-4V introduces additional overhead, which is not evaluated in the related experiments.\n\n[1] Zhang, Zhixing, et al. \"Sine: Single image editing with text-to-image diffusion models.\" CVPR 2023. \n[2] Kawar, Bahjat, et al. \"Imagic: Text-based real image editing with diffusion models.\" CVPR 2023." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "See weaknesses." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper introduces a novel evaluation metric for text-guided image editing that balances both preservation of the source image and modification toward the target text. It demonstrates remarkable improvement in alignment with human evaluators on diverse editing scenarios such as object, attribute, style alteration compared to all other existing metrics. Moreover, the metric is applicable to personalized generation, DreamBooth dataset, where the objective is to identify the source object in provided image, and generate into a completely novel context. This shows the flexibility of AugCLIP, that seamlessly apply to variety of editing directions. Notably, the metric excels in identifying minor differences between the source image and the edited image, showing superb ability in complex image editing scenarios such as MagicBrush. The major contributions are summarized as follows.\n- This paper is the first to point out CLIPScore’s reliability in text-guided image editing, as it frequently exhibits a bias towards modification rather than preservation and focuses on irrelevant regions.\n\n- This work proposes AugCLIP, a metric for image editing by automatically augmenting descriptions via LLM and estimating a balanced representation of preservation and modification, which takes into account the relative importance of each description.\n\n- In the experimental evaluations, AugCLIP demonstrates a significantly high correlation with human evaluations across various editing scenarios, even in complex applications where existing metrics struggle." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces AugCLIP, a novel evaluation metric for text-guided image editing that balances both preservation of the source image and modification toward the target text. By leveraging a multi-modal large language model to extract fine-grained visual attributes and applying a hyperplane-based optimization approach, AugCLIP estimates a representation of a well-edited image that closely aligns with human evaluators’ preferences. Extensive experiments across five benchmark datasets demonstrate AugCLIP’s superior alignment with human judgments compared to existing metrics, particularly in challenging editing tasks. Consequently, AugCLIP offers a significant advancement in the evaluation of textguided image editing, providing a more nuanced and reliable approach for assessing modifications while maintaining core image attributes. This metric holds promise for broader applications in personalized image editing and other vision-language tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Overall this work makes an interesting and meaningful observation about the widely used CLIPScore metric, however there are still some concerns:\n- **Discussion on broader indicators.** This work highlights the problem of CLIPScore in the problem analysis in Section 3. Do other quantitative indicators such as FID and LPIPS have similar problems? Please give a more comprehensive analysis.\n- **Suitability for complex editing instructions or tasks.** There are many kinds of image editing tasks, including global editing such as style editing rather than just local editing. How does AugCLIP perform in this case?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024augmentationdriven,\ntitle={Augmentation-Driven Metric for Balancing Preservation and Modification in Text-Guided Image Editing},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=08FCLXDY3S},\nnote={under review}\n}" }, "abstract": { "value": "The development of vision-language and generative models has significantly advanced text-guided image editing, which seeks \\textit{preservation} of core elements in the source image while implementing \\textit{modifications} based on the target text. However, in the absence of evaluation metrics specifically tailored for text-guided image editing, existing metrics are limited in their ability to balance the consideration of both preservation and modification. Especially, our analysis reveals that CLIPScore, the most commonly used metric, tends to favor modification, resulting in inaccurate evaluations.\nTo address this problem, we propose \\texttt{AugCLIP}, a simple yet effective evaluation metric that balances preservation and modification. \n\\texttt{AugCLIP} begins by leveraging a multi-modal large language model (MLLM) to augment detailed descriptions that encapsulate visual attributes from the source image and the target text, enabling the incorporation of richer information. Then, \\texttt{AugCLIP} estimates the modification vector that transforms the source image to align with the target text with minimum alteration as a projection into the hyperplane that separates the source and target attributes. Additionally, we account for the relative importance of each attribute considering the interdependent relationships among visual attributes. Our extensive experiments on five benchmark datasets, encompassing a diverse range of editing scenarios, demonstrate that \\texttt{AugCLIP} aligns remarkably well with human evaluation standards compared to existing metrics. The code for evaluation will be open-sourced to contribute to the community." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "evaluation metric", "text-guided image editing", "multi-modal representation" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/094577075cac891aa6b84345f9414fba7122b758.pdf" }, "presentation": null, "primary_area": { "value": "other topics in machine learning (i.e., none of the above)" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/c75a814bb45dad0ea12727675816e434503a14d5.pdf" }, "title": { "value": "Augmentation-Driven Metric for Balancing Preservation and Modification in Text-Guided Image Editing" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. Is it possible to extend CLOVER to improve performance on tasks that involve reasoning with data formats beyond natural language, such as mathematical equations or visual reasoning tasks?\n2. Can the authors provide more insights into how CLOVER compares with CoT-based methods designed for improving implicit reasoning of LLMs?\n3. Why not test CLOVER on a wider range of language models to assess its generalizability?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The paper presents a new approach by breaking down compositional translation into smaller steps and combining it with verification for logical reasoning tasks, leading to improved performance in neurosymbolic translation.\n- The experimental results are robust using GPT4-o, showing improvements over other methods across multiple benchmarks.\n- The authors propose two SAT-based first-order logic verification algorithms for selecting a sample from LLMs' logical code generations." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces CLOVER, an approach designed to enhance the translation of natural language logical problems into logical code, thereby improving the performance of language models on logical reasoning benchmarks. CLOVER achieves this by compositional translation of natural language into first-order logic and verification of logical semantics. The method involves parsing natural language sentences into logical dependency structures, translating these into first-order logic formulas, and employing verification algorithms to ensure accuracy. The authors demonstrate CLOVER's effectiveness on seven logical reasoning benchmarks, showing it outperforms previous neurosymbolic approaches." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The approach is primarily applicable to problems that can be represented in a SAT solver, limiting its generalizability to other reasoning datasets, such as those involving mathematical equations or visual components, e.g., MATH dataset.\n- The core idea of breaking down tasks into subtasks and using multiple samples and tests (e.g., verification, self-reflection, deterministic tests) to select the best generation is not novel.\n- The paper lacks comparison with chain-of-thought (CoT) based methods designed to improve implicit reasoning of language models, as in \"reliable reasoning beyond natural language\". These methods help the model extract information that is implied but not directly stated by interleaving natural language comments with logical code, and can alleviate the translation bottlenecks identified.\n- The paper only reports results using one language model, making it unclear if the method would improve performance across different models and weakening the experimental results." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "1. line 115: \"To save computational cost, we compare each one of logically equivalent formulas\". You probably mean to \"compare each logically equivalent formula\". How can this save computational cost?\n\n2. Line 149: how to read this formula in natural language? \n\n3. What is the output for the sentence \"A barber shaves all who do not shave themselves.\"? \n\n4. How are \"Declarations\" created? \n\n5. How to decide a sentence not fit for your system? (or how to decide an unintended input sentence?)" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "It is reasonable to improve the translation quality by decomposing a complex sentence into several shorter sentences. Using SAT solvers certainly improve the quality." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Authors propose a novel method of using LLMs to translate natural language descriptions into a set of first-order logical forms. This novel method decomposes this challenging task into two steps. The first step is to translate a long and complex sentence into a number of short sentences, the second step is to translate each short sentence into simple first-order logical forms and the connections between/among these short sentences into corresponding logical connectors. Experiments on seven benchmark datasets greatly outperform current SOTA level." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Not all natural language sentences can be translated to first-order logic forms. Authors did not discuss what sentences cannot be translated. \n\nAuthors use symbolic SAT solver in evaluating and selecting correct first-order logical forms. This limits the method only for the case where SAT solvers work. \n\nTheoretically, the meaning of natural language is not logical formula. This work is valued within fixed benchmark datasets. \n \nThe formalism of the paper is not easy to read." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- I'm curious why the execution rate increases when using CLOVER. As I read the methods section, it looked like CLOVER primarily helps with execution accuracy, but I didn't see much about how it would help repair/fix/generate better code for the SAT solver. \n\n- It's reported that \"CLOVER’s errors are primarily caused by preprocessing and other errors, which takes 78.6% of the total errors\", do you have examples of this? Is this an error in the accumulation stage? I think the paper does a great job of explaining where Logic-LM fails and why CLOVER is needed, but I think expanding on CLOVER errors is just as important to show where researchers can look next.\n\n- How much of the performance gain seen in CLOVER is due to a higher execution rate (runnable code)? I think expanding on how the metrics in Table 2 are computed would be helpful. For example is `Execution Acc = (correct_and_executable_programs / all)` or `Execution Acc = (correct_and_executable_programs / executable_programs)`. The latter, I think, helps distinguish if you are generating better executable problems or if you are only improving the execution rate (which maybe there is a simple fix to Logic-LM to help it create better executable problems)?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The results are promising; the authors do a fantastic job of motivating their new algorithm CLOVER by showing common failures of previous methods like Logic-LM, then show that their method fixes many of these errors (leading to the performance boost reported in Table 1).\n\n- The method here is pretty novel. Breaking down sentences into atoms isn't too novel, but I haven't seen someone decode them all individually and progressively (tapping into the auto-regressive natural of LMs) to improve the performance of the translation. The verification algorithms seem pretty intuitive (although they are described complexly), but again, despite being intuitive, I think they are fairly novel as well." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Introduces a new algorithm, CLOVER, for solving logic questions in natural language, specifically by addressing the challenges in parsing natural language into the correct first-order logic so that an SAT solver can determine the answer. To do this, the paper proposes translating the question into smaller pieces that accumulate based on how each logic unit in natural language relates to other logic units until the resulting sentence is the final sentence that needs to be tested. Each accumulation from the previous step, including the final sentence, is translated into first-order logic, then the paper introduces two novel verification methods that check if the translations are valid and if there are contradictions in the final translation. The paper shows that with this accumulation strategy with their two verifiers, their method can outperform all baselines (including Logic-LM, a baseline that similarly translates sentences into FOL for SAT solvers) on common logic datasets like AR-LSAT, FOLIO, and ProofWriter." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The ablations in Table 3 need to be explained more clearly and then discussed. What is \"Is Clover?\" and why is the simplest ablation (no clover, direct translation, no verification / essentially none of the new things introduced in this paper) outperforming Logic-LM on AR-LSAT by 10.9% already from Table 1? Does this mean that your direct translation prompt already improves over complex algorithms like Logic-LM? If so, this deflates the papers impact, so it should be addressed (it's possible I am missing something, but others will catch this too, so it's best to explain it away.)\n\n- I believe the paper would benefit greatly from expanding on the models being evaluated; right now, only GPT-4o and GPT-4o-mini are evaluated. Showing that CLOVER consistently outperforms the baseline methods across model classes would improve the impact of this work.\n\n- (minor point) There is no discussion of inference-time compute costs for CLOVER vs. the other baselines. I imagine the inference cost is significantly higher, but I am unsure how much. Is this negligible compared to Logic-LM? Is there a way to compare CLOVER with baselines that use the equivalent amount of compute during inference? I think much of this point could be explained away with a textual justification (i.e., this isn't possible, or the compute costs are nearly equivalent, etc.), but I do think it should be mentioned.\n\n- (minor point) Clarity in section 3 could be improved. I would use the example in Figure 2 to clearly define each variable mentioned in the text to help readers follow your algorithm. For instance, defining with x^prep, T, phi_k, the mapping NL(phi), etc., with values from Figure 2 would help readers follow significantly. This could also be done in Figure 2 if you mark which parts of it are which variables. The text gets very dense with variables that are derived from other variables quickly; having these concrete instantiations really helps." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We introduce CLOVER, a neurosymbolic approach that enhances complex logical reasoning in large language models by compositional translation of natural language into first-order logic and verification of logical semantics." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024divide,\ntitle={Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=09FiNmvNMw},\nnote={under review}\n}" }, "abstract": { "value": "Complex logical reasoning tasks require a long sequence of reasoning, which a large language model (LLM) with chain-of-thought prompting still falls short. To alleviate this issue, neurosymbolic approaches incorporate a symbolic solver. Specifically, an LLM only translates a natural language problem into a satisfiability (SAT) problem that consists of first-order logic formulas, and a sound symbolic solver returns a mathematically correct solution. However, we discover that LLMs have difficulties to capture complex logical semantics hidden in the natural language during translation. To resolve this limitation, we propose a Compositional First-Order Logic Translation. An LLM first parses a natural language sentence into newly defined logical dependency structures that consist of an atomic subsentence and its dependents, then sequentially translate the parsed subsentences. Since multiple logical dependency structures and sequential translations are possible for a single sentence, we also introduce two Verification algorithms to ensure more reliable results. We utilize an SAT solver to rigorously compare semantics of generated first-order logic formulas and select the most probable one. We evaluate the proposed method, dubbed CLOVER, on seven logical reasoning benchmarks and show that it outperforms the previous neurosymbolic approaches and achieves new state-of-the-art results." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Logical Reasoning", "Large Language Models", "Neurosymbolic Approaches", "Semantic Decomposition", "Formal Language Verification" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/34452baa1c927a1217586b4f505eb5345c199dae.pdf" }, "presentation": null, "primary_area": { "value": "neurosymbolic & hybrid AI systems (physics-informed, logic & formal reasoning, etc.)" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. How does the use of gradient ascent and descent for patch derivation differ from recent work in unlearning?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The problem addressed—post-hoc safety alignment—is important for ensuring that LLMs behave safely in real-world applications.\n2. The empirical evaluation and ablations are fairly comprehensive across different LLM backbones and benchmarks.\n3. The method shows some promise in balancing safety with utility preservation compared to existing baselines." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents a method called SafePatching for post safety alignment (PSA) of large language models (LLMs). The authors claim that SafePatching addresses three PSA objectives: safety enhancement, over-safety mitigation, and utility preservation." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The proposed approach seems to be largely composed of a series of straightforward adaptations or incremental improvements on recent work. For instance, the use of gradient ascent and descent techniques for deriving safety and over-safety patches is largely an adaptation of existing machine unlearning methods described in the paper, rather than a truly novel contribution. The concept of patching the difference set of important parameters between safety and over-safety patches is perhaps the most novel aspect. However, it's still a relatively straightforward extension of existing ideas in parameter importance and model merging.\n2. While the proposed approach does demonstrate that it is the only one to improve safety, over-safety, and utility over the backbone, in many cases, it performs significantly worse than the baselines for a particular safety or over-safety benchmark. Moreover, the safety and over-safety improvements over the backbone model are quite marginal in some cases. This highlights that there is more work to be done in effectively controlling the balance between safety enhancement and over-safety mitigation than the approach in its current state." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Would you please address the concerns in weakness?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "•\tThe idea is straightforward.\n•\tThe experiments are extensive." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a post safety alignment method which merges two models post-trained on harmful data with gradient ascent and descent respectively. The post-trained and merged model preserves a balance on safety, over-safety mitigation, and utility preservation." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "•\tThe paper lacks the comparision of external safeguards methods such as OpenChatKit and NeMo guardrails that are known to handle over-safety issues. Would these external safeguards methods also achieve the three objectives proposed in the paper?\n•\tThere are a few hyperparameters in equation 7&8, such as a, b, \\alpha, \\beta. How you set these parameters? In Table 3, merging methods like the task arithmetic and TIES-merging do not have big differences compared to the intersect patch. Would the benefit comes from your hyperparameter selection?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "Yes, Discrimination / bias / fairness concerns", "Yes, Privacy, security and safety" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* Given that only the AdvBench dataset was used to evaluate SAFEPATCHING, how does the method perform across other safety-related datasets? Could testing with a broader range of harmful data enhance our understanding of its transferability to diverse safety scenarios?\n* Since the authors did not specify whether they directly used the fine-tuned Longformer model from Wang et al. or performed additional fine-tuning, what impact might this setup have on the accuracy and reliability of the judgment model in this experiment?\n* Could a deeper explanation of these aspects clarify the novelty and rigor of the proposed approach in Section 3.3?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "* The paper proposes a new method named SAFEPATCHING to address the limitations of existing methods on post-safety alignment for LLMs, such as over-safety issues and high cost.\n* The paper presents experimental results and comparisons with state-of-the-art methods to demonstrate the effectiveness of SAFEPATCHING and uses multiple open-source datasets on safety, over-safety, and utility for a comprehensive evaluation. Besides, this paper has interesting findings on the distribution of the most important parameters for safety and over-safety, providing future research directions for the community." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a novel post-safety alignment (PSA) method, called SAFEPATCHING, which aims to address safety, over-safety, and utility issues in large language models (LLMs). In this paper, the authors develop a two-stage PSA framework, which applies distinct safety patches to the backbone LLM based on harmful data to improve safety and reduce over-safety, meanwhile, maintaining the utility capability of the LLM. The experiment shows that SAFEPATCHING achieves more effective and efficient PSA compared to baseline methods across four aligned LLMs" }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* Lack of justification in Sec. 3.3 controllable patching. The authors may want to highlight the novelty of their tool and the rigor of their method. Currently, it appears that the approach relies on the SNIP score proposed by Lee et al., as well as model merging methods by Yu et al. and Hui et al., without a thorough explanation of the unique contributions or advancements made in this work.\n* Although the authors conducted an excessive experiment to show the effectiveness of SAFEPATCHING, several concerns existed in the settings.\n * The study evaluates SAFEPATCHING using only a single harmful dataset, AdvBench, which may not adequately demonstrate the method's transferability across different safety scenarios. Given the extensive range of safety categories and perspectives, it's essential to assess whether a backbone LLM patched using AdvBench can maintain its effectiveness on other datasets representing diverse types of harmful content.\n * The authors did not specify how they fine-tuned the Longformer-based judger. Wang et al. used annotated data generated through human labor to fine-tune their Longformer model. It remains unclear whether the fine-tuned model from Wang et al.'s work was directly utilized in this experiment or if further adjustments were made. Clarification on this point would provide a better understanding of the model’s setup and any adaptations relevant to this study.\n\nYu, Le, et al. \"Language models are super mario: Absorbing abilities from homologous models as a free lunch.\" Forty-first International Conference on Machine Learning. 2024.\n \nHui, Tingfeng, et al. \"HFT: Half Fine-Tuning for Large Language Models.\" \n \nWang, Yuxia, et al. \"Do-not-answer: A dataset for evaluating safeguards in llms.\"" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "- Potential for Misuse and Safety Bypass\n\nThe SafePatching framework’s dual-patch approach is designed to mitigate over-safety, allowing the model to respond to benign prompts with sensitive keywords. However, this opens up a risk of misuse if bad actors attempt to exploit this flexibility to bypass safety mechanisms deliberately." }, "flag_for_ethics_review": { "value": [ "Yes, Privacy, security and safety", "Yes, Potentially harmful insights, methodologies and applications" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "In the SafePatching framework, Eq (1) and Eq (2) are designed to achieve two opposing objective by applying gradient-based updates in opposite directions on the same harmful dataset. Could you please clarify and elaborate how they are implemented given a harmful dataset?\n\n\nIn SafePatching, what requirements should a harmful dataset fulfill? For example, are there specific expectations concerning its size, diversity, or other characteristics? Additionally, are these requirements realistic for SafePatching's application in real-world scenarios?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "+ The proposed approach is easy to understand, logical, and appears to be effective.\n\n+ It addresses a significant and timely problem.\n\n+ The paper is overall well-written.\n\n+ The unlearning and fine-tuning techniques used in SafePatch are not new, the originality comes from considering dual patching at the same time.\n\n+ The paper includes an extensive set of experiments." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces a SafePatching framework to improve the safety of large language models (LLMs) while maintaining their utility. The major contribution is two types of safety patches. \n- The safety enhancement patch utilizes gradient ascent on harmful data to train the model to avoid generating unsafe responses. It effectively helps the model \"unlearn\" unsafe behaviors by adjusting the model parameters to minimize the risk of producing harmful content. \n- The over-safety mitigation patch, developed through gradient descent, is designed to prevent the model from being overly cautious. It fine-tunes the model to ensure it does not overly restrict or reject benign inputs that might superficially appear sensitive or risky. \n\n The approach is tested across multiple LLMs, showing better performance in reducing harmful outputs, handling over-safety, and preserving utility compared to several existing methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Limited Novelty in Core Techniques\n\nWhile the dual-patching approach is innovative in combining safety enhancement with over-safety mitigation, the core methods (e.g., gradient ascent and descent on harmful data) rely heavily on existing unlearning and fine-tuning techniques.\n\n- Clarity on Practical Deployment\n\nThe paper would benefit from more actionable details regarding the real-world deployment of SafePatching, especially the requirements on the harmful data set.\n\n- Stability of SafePatching Approach\n\nSafePatching's dual-patch integration requires careful parameter tuning, especially with the two gradient-based patches potentially introducing conflicts within the model. The process of managing these interactions, although effective, may lack robustness or generalizability across different architectures or types of prompts." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024towards,\ntitle={Towards Comprehensive and Efficient Post Safety Alignment of Large Language Models via Safety Patching},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=09JVxsEZPf},\nnote={under review}\n}" }, "abstract": { "value": "Safety alignment of large language models (LLMs) has been gaining increasing attention. However, current safety-aligned LLMs suffer from the fragile and imbalanced safety mechanisms, which can still be induced to generate unsafe responses, exhibit over-safety by rejecting safe user inputs, and fail to preserve general utility after safety alignment. To this end, we propose a novel post safety alignment (PSA) method to address these inherent and emerging safety challenges, including safety enhancement, over-safety mitigation, and utility preservation. In specific, we introduce \\textsc{SafePatching}, a novel framework for comprehensive and efficient PSA, where two distinct safety patches are developed on the harmful data to enhance safety and mitigate over-safety concerns, and then seamlessly integrated into the target LLM backbone without compromising its utility. Extensive experiments on four representative aligned LLMs, including LLaMA-2/3, Gemma and Mistral, show that \\textsc{SafePatching} achieves a more comprehensive and efficient PSA than baseline methods. It even enhances the utility of the backbone, further optimizing the balance between being helpful and harmless in current aligned LLMs. Also, \\textsc{SafePatching} demonstrates its superiority in continual PSA scenarios. \\textcolor{red}{WARNING: This paper may contain content that is offensive and harmful.}" }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Post Safety Alignment", "Large Language Models", "Jailbreak Defense", "Over-Safety Mitigation" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/b28f0d1191a8411bdc4d03df6e9a2f113daa2a14.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/5eec85122a6ffebd9e8aacea99e1f803e3556094.zip" }, "title": { "value": "Towards Comprehensive and Efficient Post Safety Alignment of Large Language Models via Safety Patching" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "* Did you do any finetuning over the used models, notably LLMs or are you using frozen models ?\n* Why cannot you compare to any existing baselines ?\n* Have you optimized the creation of your pipeline using these 5 kaggle competitions, or have you left out some of them, to evaluate on competitions you did not know at design time ?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "**Interesting problem.** With the LLMs (+RAG) becoming mature, the open source study of their integration into broader tools that can directly be applied to data science tasks, is the natural next step.\n\n**Overall good presentation.** Even if some details are lacking to grasp the authors' exact contribution (notably in the figures), the overall presentation clearly demonstrates the problem and the approach set up to tackle it. \n\n**Interesting metrics and ablation studies.**" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces AutoKaggle, a pipeline to automatically solve Kaggle Competitions. The authors use 5 subparts in a row: a reader, a planner, a developer, a reviewer, and a summarizer. They use LLMs with RAG to develop code-based solutions, with code running, units tests. They evaluate their method on 5 Kaggle competition benchmarks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "**Lacking evaluation.** The evaluation is lacking comparison to existing AutoML baselines (*e.g. [1]) or explanations on why the authors are not comparing their method to any existing solution. If running such comparison is not possible at all, then the authors should provide explanations on why this is not feasible. \nWhile detailed reports are provided on their methods and the different components, as this works apply existing techniques, its evaluation is its core contribution. \nThe authors should report (at least) the standard deviation, but e.g. violin plots to compare AutoKaggle's results of other kaggle competitors could help clearly situate where this automatic pipeline stands.\n\n**Evaluation on a (previously) unknown dataset.** It seems that AutoKaggle has been designed to solve these datasets, so one cannot evaluate how much this method would transfer to another, previously unknown dataset.\nIt would be nice to provide the reader with how much out of the box your method is, maybe with a user study. It seems like its your core contribution, so having independent people trying AutoKaggle and commenting on how easy the setup and interaction is on a left out dataset would help people looking for such solutions.\n\n**Figure 2 could be improved.** The figure could be split to separate the overall pipeline from details on some of its components. Most importantly, what part is using an LLM, what part is using a human expert ? This figure represents 70% of what the reader is looking for, it should provide first the overall intuition, and then enough details on specific core components that you want to highlight.\n\n**You related work section is actually a background section.**\nYour current related work covers some domains that are integrated within AutoKaggle. It thus feels more like a related work of your background section (what AutoKaggle builds upon). Is there any *e.g.* AutoML method that you can compare to ? Any method that addresses the same issue ?\n\n\n[1] https://github.com/automl/CAAFE" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please refer to the questions in Weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- AutoKaggle introduces a tailored phase-based workflow with multi-agent collaboration specifically designed for data science competitions. The system’s demonstrated high average completion rate and competitive ranking in Kaggle highlight its effectiveness, particularly in tabular classification and regression tasks, showing its strength in handling structured data challenges.\n\n- AutoKaggle empowers the Developer agent to perform iterative debugging and unit testing, bolstering the robustness of code generation. Additionally, the integration of a comprehensive machine learning tools library improves the system's efficiency and accuracy, making it better suited for tackling complex Kaggle competitions" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents AutoKaggle, a multi-agent framework specifically designed to handle the complexities of Kaggle data science competitions. The framework organizes the competition workflow into six distinct phases—background understanding, exploratory data analysis, data cleaning, in-depth exploratory analysis, feature engineering, and model development and validation—allowing agents to work systematically through each stage. Key agents, including Reader, Planner, Developer, Reviewer, and Summarizer, collaborate within this structure, with iterative debugging and unit testing to ensure robustness and accuracy in code generation. AutoKaggle integrates a machine learning tools library to streamline tasks, enhance code reliability, and provide users with educational insights through comprehensive reports at each phase. Evaluated across multiple Kaggle competitions, the framework achieved an average completion rate of 83.8% and ranked in the top 42.8% in Kaggle." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Limited novelty. While the paper addresses data science problem-solving using LLM-based agents, it lacks a clear description of the specific challenges it intends to solve that existing methods have struggled with. Extending from a single-agent to a multi-agent system is insufficiently justified in this field, as the necessity and performance gains of such an approach are not clearly demonstrated. Existing works, as mentioned in the introduction, have also tackled similar problems with LLM-based agents, questioning the incremental contribution of AutoKaggle.\n\n- Multi-agent system design. The multi-agent system, including agents like Reader, Planner, Developer, Reviewer, and Summarizer, is insufficiently explained in terms of its collaborative structure. It is unclear whether these agents operate in an assembly-line fashion or if they engage collectively in each phase under the \"Cooperative Engagement\" label in Figure 1. Further clarification on their integration and interdependence within each workflow phase is needed.\n\n- Role clarity of Planner and Summarizer. Given AutoKaggle’s sequential, phase-based workflow, the necessity of a Planner agent is ambiguous. Can you quantify the contribution (such as on completion rates or error reduction) of this Planner agent in your system? Similarly, the Summarizer’s role in contributing to critical performance metrics such as completion rate or Best Normalized Performance Score, is not explicitly justified, leaving its impact on performance uncertain.\n\n- Unit Test and Debugging. Dose the Developer agent generate dataset-specific unit tests that align with each unique code snippet or not? How the Developer agent adjusts unit tests based on code variations to ensure logical consistency and accuracy across different tasks? \n\n- Lines 275-276 mention the importance of detecting logical errors in code, yet the method for achieving this is underexplored. Can you explain more details about detecting the logical error? More detail is needed on how logical errors are detected and avoided, as conducting exploratory data analysis or statistical checks after data cleaning or feature engineering alone may be insufficient. \n\n- Table 2 illustrates the system's performance across different debugging attempts (DT), showing how increased debugging impacts metrics like Completion Rate (CR) and Comprehensive Score (CS). The data indicate that both CR and CS improve as DT rises, reflecting enhanced task completion and accuracy with more debugging opportunities. What the 'performance plateaus' mean in line 524-525?\n\n- The paper does not provide information on the cost of running AutoKaggle, which is essential for evaluating its performance and practical applicability. It's benifit to provide cost and total runtime to understand the performance. \n\n- The chosen baselines are not entirely convincing. Recent similar works, AIDE[1] and MLE-Agent[2] have shown remarkable capability in Kaggle competition settings. A comparative analysis with these recent works, particularly focusing on AutoKaggle’s unique advantages in effectiveness, efficiency, or other performance metrics, would highlight its distinct contributions to the field.\n\n- A broader evaluation across various task types such as time series prediction, image classification, and text classification, are necessary, as these are critical and challenging categories in Kaggle competitions. The current experiments focus primarily on tabular datasets, leaving it unclear whether AutoKaggle is capable of handling more complex, domain-specific tasks. Can AutoKaggle complete such tasks? \n\n- What is the requirement of the LLM? Can AutoKaggle works well with gpt-3.5 or other open-sourced models?\n\n\n[1] AIDE: the Machine Learning Engineer Agent(https://github.com/WecoAI/aideml)\n\n[2] MLE-Agent: Your intelligent companion for seamless AI engineering and research (https://github.com/MLSysOps/MLE-agent)" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "~L140, you say that CoT improves reasoning at the expense of introducing hallucinations. Is there any evidence that CoT makes models any more or less likely to hallucinate?\n\n~L141, you say that the ReAct paradigm addresses hallucinations - that's not my understanding of what ReAct does or how it works, my understanding is that it combines thoughts and actions, yes, but that this has nothing to do with hallucinations or refining outputs.\n\n~L360: What is the difference between \"Success - Non-compliant\" and \"Success - Compliant\"?\n\n~L403: What's the justification / motivation for the complex / compound \"Comprehensive Score\"? How does it compare to other measures, what specifically does it achieve or avoid?\n\n~L431: could you say more about this \"strong baseline\" - I don't understand its construction.\n\nIf adding FE tools drops performance because FE adds too much complexity, then why does \"All tools\" (which presumably includes FE tools) recover this performance?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "A system which can score 43% on Kaggle leaderboards is a significant milestone on the path to automated coding and datascience. Additionally, since many challenges which such a system would face would also arise in more general task-completion (e.g. long-term planning, establishing coherency, managing context, preventing execution from looping) and so would transfer to improve AI agents in general.\n\nGreat collection of Classic and Recent challenges, and baselines seem reasonable (though see my Q about the Strong Baseline).\n\nIt's helpful to have this variety of scores (though see my Q about CS).\n\nArchitecture is clearly laid out, and the paper is overall very easy to read.\n\nClear exploration and explanation of the underlyring readon why the feature-engineering tools reduce the framework's score (many features, leading to more complexity than the agents can handle)." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a scaffolding framework which uses LLMs to create a multi-agent system, used to attempt Kaggle problems. They use a \"phase-based\" multi-agent approach, together with a library of hand-crafted ML tools, and extensive hand-crafted unit-tests tailored to the Kaggle problems.\n\nApplying this framework to 8 Kaggle problems (4 pre-GPT-4 training cut-off, 4 afterwards), they achieve a significant solve rate, and an average of 42% on the Kaggle leaderboard.\n\nThe paper also explores ablation of various modules (various tools, and the unit-testing module)." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Whenever CoT is used as an interpretability tool, I think it's always wise to mention unfaithfulness e.g. https://arxiv.org/abs/2305.04388\n\nThere are two places where a long list is hard to read:\n#1 ~L78: AutoKaggle integrates a comprehensive machine learning tools library, covering three core toolsets: data cleaning, feature engineering, and model building, validation, and prediction.\n\n#2 ~L186: The data science process is divided into six key stages: understanding the\nbackground, preliminary exploratory data analysis, data cleaning, in-depth exploratory data anal-\nysis, feature engineering, and model building, validation, and prediction\n\nPerhaps \"model-building, -validation, and -prediction\" would be easier to read.\n\n~L146: I'm surprised not to see mentioned what seems to me to be the main thing underlying the motivation of multi-agent systems: finite context length, requiring summarisation and specialisation.\n\nIt's not clear how much of the headline 43% on the leaderboard is down to the skill of the human-in-the-loop, which severely undermines the claim. Without a comparison to how well the human takes unassisted (in terms of success rate or time taken), or to how well AutoKaggle performs without HITL, it's impossible to reliably state how effective the framework is.\n\nUnspecified HITL also undermines the various claims of a \"fully automated framework\" (e.g. L175)\n\nNot much detail on these unit tests. Who writes them? What's the coverage like? Are there any guarantees? If (as I suspect) the \"meticulously designed\" unit tests are written by humans, then we have a similar situation as with the unspecified human-in-the-loop: the framework is not \"fully automated\", and it's impossible to rigorously determine how much effect the human hand-holding has on the framework's suggess. This should, at minimum, be clearly, explicitly and boldly acknowledged.\n\nAdditionally, it is unclear to me how much of the ML-tools library was developed alongside particular Kaggle Competition attempts. If the tools were developed on a case-by-case basis, to address hurdles found in the challenge, then there is significant data leakage from the evaluation dataset to the framework, leading to overfitting to the competitions chosen during development, and much of the headline 43% comes from tools handcrafted by human developers on a case-by-case basis. For a fair validation of how well this framework performs in \"fully automated\" mode, the library would need to be \"frozen\" while the framework was tested on a held-out set of Kaggle Competitions.\n\nVery minor point: ~L350, I agree that there is a risk of data leakage for competitions from before Oct '23, however to say that GPT-4o's training data includes Classic Kaggle is an assumption: better to say simply that there is a risk of data leakage.\n\nIf you're considering data leakage, it would be worth flagging that the 42% includes Classic problems: using only the newer problems, performance is slightly below human average." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose AutoKaggleMaster, a robust and user-friendly framework that solves Kaggle problems through a multi-agent collaborative system." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024autokaggle,\ntitle={AutoKaggle: A Multi-Agent Framework for Autonomous Data Science Competitions},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=09LEjbLcZW},\nnote={under review}\n}" }, "abstract": { "value": "Data science competitions on Kaggle, which represent real-world programming challenges, require sophisticated problem-solving approaches. While LLM-based agents demonstrate potential in various fields, their application to data science tasks often falls short due to difficulties in adapting to data changes in multi-stage reasoning and the need for precise reasoning. To address this, we propose AutoKaggle, a robust and user-centric framework that solves Kaggle problems through a collaborative multi-agent cooperative system. AutoKaggle implements an iterative development process that combines code interpretation, debugging, and comprehensive unit testing covering over 30 tests, ensuring code correctness and quality through LLM-based evaluation. It prioritizes user experience by generating detailed reports that elucidate feature engineering processes, data transformations, model selection criteria, and the reasoning behind each decision. It offers customizable workflows, allowing users to intervene and modify each stage of the process, thus combining the advantages of automated intelligence with human expertise. Additionally, we build a universal data science tool library, including carefully verified functions for data cleaning, feature engineering, and modeling, which form the foundation of this solution. We evaluate the framework on 8 carefully selected Kaggle competitions, achieve 83.8\\% in average completion rate and 42.8\\% average rank in Kaggle." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "large language models", "language agents", "multi-agent" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/06db613e04c157c36722c2631ca958f7f587be75.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "AutoKaggle: A Multi-Agent Framework for Autonomous Data Science Competitions" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "In Eq. 7, L_richness. It would be useful to give dimensions of X and F. Possibly also to rewrite the equation to make the way this is calculated easier to understand.\n\nIn Sec 3.3. - How is noise Z added? Is it sampled once and used for all blocks?\n\nIn Eq. 11, which features are max(s) and min(s) calculated from in the normalization?\n\nWhy no M3DM comparison or comparison on MVTec3D or on Real3DAD that have been published and are more widely cited?\n\nThe BTF method achieves an extremely low AUROC score on the ICD dataset showing a strong correlation between the anomaly score and the normality of the example which may be interesting and should be commented on given that the dataset is one of the contributions.\n\nThe discussion of the experimetnal results could be expanded.\n\nWhy are the results on the ICD dataset relatively low in terms of the AUROC scores?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "- Interesting method that aims to improve existing handcrafted features which seems novel.\n- The proposed method achieves state-of-the-art results on the AnomalyShapeNet dataset and on the newly proposed ICD dataset.\n- Most sections of the paper are well written and easy to follow despite the proposed method being constructed of several components." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose a novel method that uses a noise prior to learn to improve the features of a handcrafted descriptor FPFH. The FPFH features are reformulated through a series of Information Gain blocks that attempt to extract useful information from noised FPFH features thus decoupling the noise from the useful information contained within the features. The extracted features are then used to create a memory bank which is used at inference for anomaly score estimation. A packet downsampling process is also proposed, which is a Mahalanobis distance-based greedy coreset sampling mechanism that better samples features in cases where the observed class is composed of several subclasses.\n\nThe authors also propose a new dataset, ICD, where each class is composed of several subclasses providing a unique challenge for 3D anomaly detection methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Some implementation details, such as how noise is injected and sampled during the pretraining phase, could be included as it would improve clarity for the reader. Some implementation details are included in the supplementary but could be moved to the main paper.\n\nThe results of the comparison of the proposed method to related works could be discussed in more detail. Currently the results on AnomalyShapeNet and ICD are only briefly listed in Section 4.3, however no discussion of the results is given.\n\nOn the ICD dataset the current SOTA on AnomalyShapeNet is not evaluated (R3D-AD) and the second best method is a vanilla PatchCore using FPFH features which generally does not achieve SOTA results on 3D anomaly detection benchmarks. Given that the ICD dataset is one of the claimed contributions of this paper the evaluation should be more thorough and the discussion of the results more detailed.\n\nThe ablation study is done on the newly proposed ICD where the performance is very low (0.6 AUC). This makes it difficult to really evaluate the components of the method since most anomalies are already missed and the difference between most experiments is less than 1% AUROC.\n\nOverall I believe the experimental section is the most lacking. There is a lack of discussion of the results on both the AnomalyShapeNet and the ICD dataset. Additionally, the evaluation on the ICD dataset could be more thorough. Methods that are included in the AnomalyShapeNet experiments are not included in the ICD experiments. The results are not properly discussed. Only image-level AUROC is used for the evaluation in Section 4.3 but in the ablation study (Section 4.4) other metrics are also used. The ablation study should also be done on AnomalyShapeNet to get a clearer picture of the impact of each design choice." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The authors seem to achieve better performance by stacking layers of IGB and increasing the number of MLPs within them. Is this performance improvement due to increased computational complexity?\n2. In Table 3, the results without using IP and IGB appear to be better than those with IP and two layers of IGB. Please explain the effectiveness of IGB and IP.\n3. The comparison methods in Table 1 differ from those in Table 2. It seems that the experimental results of CPMF, IMRNet, and R3D-AD on the ICD dataset are missing in Table 2. It is recommended that the authors include these results to demonstrate the reliability of the experiments.\n4. The proposed dataset does not seem to have a significant advantage in terms of defect types and quantity. It appears to be a selection of a few subclasses from each category in the ModelNet dataset." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The authors propose an Information Gain Block-based Anomaly Detection method to address the issue of high intra-class variance. They introduce Rotation-Invariant Farthest Point Sampling and an Information Perfusion module composed of Information Gain Blocks. The authors incorporate noise into 3D anomaly detection to provide more distinctive feature information. Additionally, they construct the Intra-Class Diversity (ICD) 3D anomaly detection dataset" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose an Information Gain Block-based Anomaly Detection method to address the issue of high intra-class variance. They introduce Rotation-Invariant Farthest Point Sampling and an Information Perfusion module composed of Information Gain Blocks. The authors incorporate noise into 3D anomaly detection to provide more distinctive feature information. Additionally, they construct the Intra-Class Diversity (ICD) 3D anomaly detection dataset. The effectiveness of the method is validated on the constructed dataset and the ShapeNet dataset." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "See questions." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "1. The definition of 'prior noise' is missing. The authors mention 'prior noise' in the abstract and introduction but do not provide a definition, nor is it described in the methods section.\n\n2. How does the proposed method tackle the challenge of high intra-class variance?\n\n3. In Page-2 Line-77, what is the link between extracting valuable information and high intra-class variance?\n\n4. In Table 3, the unit for \"Time Cost\" needs to be provided, whether it is seconds or milliseconds.\n\n5. Are there any hyperparameters in the proposed method? Are they sensitive?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1.This manuscript proposes IGB, IGB-AD, RIFPS, IP, and PD module to enhance 3D anomaly detection.\n\n2.This manuscript introduces the ICD dataset for 3D anomaly detection. Different from existing datasets, it includes multiple sub-classes.\n\n3.The method proposed in the manuscript achieves the state-of-the-arts performance on one public dataset and the proposed dataset" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This manuscript claims that most existing 3D anomaly detection methods require the usage of registration to preprocess point clouds and exhibit high intra-class variance. To this end, it proposes IGB, IGB-AD, RIFPS, IP, and PD module to enhance 3D anomaly detection and alleviate these two challenges. Furthermore, it develops an Intra-Class Diversity (ICD) 3D dataset with multiple subclasses. Moreover, the proposed method achieves the state-of-the-arts performance on one public dataset and the proposed dataset." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1.The citation format is incorrect, with many references needing to be placed in parentheses. The authors should carefully read Section 4.1 of the Formatting Instructions for ICLR 2025 Conference Submissions.\n\n2.Lack of details on ICD datasets. Since the ICD dataset is the second contribution, the motivation for its creation should be described in the introduction section. \n\n3.The Introduction section could be better articulated. The author spends most of the Introduction describing the current issues with 3D anomaly detection but does not explain how their proposed method effectively addresses these challenges. Deeper insights need to be provided.\n\n4.In Page-2 Line-68, R3D-AD reconstructs normal samples from pseudo abnormal point clouds using a Diffusion model and cannot be categorized as a distillation method.\n\n5.Lack of experiments. 1) the ablation and comparison experiment on the proposed Rotation-Invariant Farthest Point Sampling (RFPS) 2)The performance of the proposed method on the Real3D-AD dataset." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "See the weakness." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper offers a comprehensive literature review." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes Information Gain Block-based Anomaly Detection (IGB-AD) for 3D anomaly detection to address the challenges of insufficient anomaly detection information and high intra-class variance. Overall, the writing is not clear, and the experimental results fail to demonstrate the superiority of the proposed method." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "it would be better to specify your title to include 3D anomaly detection or point cloud anomaly detection to be more specific.\n\nWhat is your definition of noise? The lack of definition makes the motivation hard for me to understand. like, \"Noise, as a prior source of information, consists of a combination of various types of data\", what is noise?\n\nWhy the teacher-student distillation networks are proposed to mitigate the effects of noise in Lines 65 and 67? I am not convinced by this claim. Like in 3DST, RD4AD, CDO, etc., is there any technique related to noise?\n\nthe description of the method is hard to understand as well. It would be better if you could improve the overview of your method a bit. Currently, I am not clear about your motivation for the framework, yet the relationships between the proposed components and the motivation are unclear.\n\nThe authors only conduct experiments on Anomaly-Shapenet and the established dataset. What about Real3D and MVTec 3D?\n\nWe can see in Table 1, that the proposed method can even perform worse than a simple baseline FPFH in some categories, which is confusing and fails to demonstrate the effectiveness of the proposed method.\n\nAlso, what about the point-level results? In Table 1 and Table 2, only object-level results are presented.\n\nThe ablation results in Table 3 fail to demonstrate the effectiveness of individual components since the variation is not significant enough. We can see that with only PD, the authors even achieve higher P-AUROC than some other variants like in rows 1, and 4 of Table 3." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024noise,\ntitle={Noise is More Than Just Interference: Information Infusion Networks for Anomaly Detection},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=09TI1yUo9K},\nnote={under review}\n}" }, "abstract": { "value": "3D anomaly detection is a crucial task in computer vision, aiming to identify anomalous points or regions from point cloud data. However, existing methods may encounter challenges when handling point clouds with high intra-class variance, especially for methods that rely on registration techniques. In this study, we propose a novel 3D anomaly detection method, termed Information Gain Block-based Anomaly Detection (IGB-AD), to address the challenges of insufficient anomaly detection information and high intra-class variance. To extract ordered features from 3D point clouds, the technique of Rotation-Invariant Farthest Point Sampling (RIFPS) is first introduced. Then, an Information Perfusion (IP) module composed of stacked Information Gain Blocks (IGB) is proposed to utilize prior noise to provide more distinguishing information for the features, where IGB is designed to utilize noise in a reverse-thinking manner to enhance anomaly detection. Finally, a Packet Downsampling (PD) technique is developed to preserve key information between multiple clusters to solve the complex downsampling situation. The main purpose of the framework is to utilize the effective information within prior noise to provide more detection criteria for anomaly detection. In addition, an Intra-Class Diversity (ICD) 3D dataset is constructed, which contains multiple categories with high class-variance. Experimental results show that the proposed IGB-AD method achieves the State-Of-The-Arts (SOTA) performance on the Anomaly ShapeNet dataset, with an P-AUROC of 81.5% and I-AUROC of 80.9%, and also gains the best performance on the ICD dataset, with an P-AUROC of 57.4% and I-AUROC of 60.2%. Our dataset will be released after acceptance." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Self-supervised learning", "Anomaly detection" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/9ae64af9f3e13bb39bc76f276404a4bb1ef10d05.pdf" }, "presentation": null, "primary_area": { "value": "unsupervised, self-supervised, semi-supervised, and supervised representation learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Noise is More Than Just Interference: Information Infusion Networks for Anomaly Detection" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": null, "comment": { "value": "We thank the reviewers for their valuable comments and we will revise accordingly." }, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": null, "primary_area": null, "questions": null, "rating": null, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": null, "summary": null, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": null, "withdrawal_confirmation": { "value": "I have read and agree with the venue's withdrawal policy on behalf of myself and my co-authors." } }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. In the introduction, the authors mention that the proposed framework is also compatible with Transformer-based LLMs, but there seems no experiments on applying the proposed method on transformer LLMs?\n\n2. What is the Merge operator in the equation (8)?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The proposed method achieve competitive results on benchmarks like Open-ended VQA and challenge sets. It outperforms LLaVA-1.5 with less training time.\n\n2. The proposed method has good intuition on how to better utilize the Mamba's efficiency with good visualizations." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper propose a new way to integrate the Mamba architecture into the multi-modal large language models (MLLM). The technical contribution include: 1. propose using visual long sequence to utilize Mamba's linear complexity. 2. design a cross-stitch scanning approach to extract and combine spatial and semantic features simultaneously. The proposed method outperforms LLaVA-1.5 with less training time and better inference efficiency, and achieve similar performance with model trained on larger dataset such as Qwen-VL." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The proposed method is likely to be dependent on vision encoders. It would be more solid if the author could conduct additional experiments on encoders other than DINOv2 + SigLIP. Also, the author does not show how proposed method perform on single vision encoder MLLM.\n\n2. There is not enough ablations experiments on the scanning orders. For example, no comparison with only using Hv1." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see weakness above." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The authors develop visual long sequences that enhance representation capabilities, ensuring more robust and detailed visual data processing.\n- The authors introduce an innovative cross-stitch scanning mechanism designed to improve the interaction between visual and linguistic data, optimizing vision-language alignment.\n- The authors present MambaVLM-a robust and streamlined MLLM framework. Their extensive testing across various benchmarks validates the effectiveness of their approach." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces MambaVLM, a novel framework that utilizes the Mamba model, a state-of-the-art selective structured state space model renowned for its linear computational complexity and its efficiency in managing long sequences. The authors enhance the Mamba model by incorporating visual long sequences and a cross-stitch scanning mechanism, specifically tailored to boost interaction and alignment between visual and linguistic data. Through extensive experiments and qualitative analyses, they establish MambaVLM not only as a powerful tool for MLLM tasks but also as a pioneering approach that sets a new benchmark for future research in the field." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The contributions are vague; it would be better to clearly summarize the contributions of this paper at the end of the Introduction. This article simply replaces the traditional MLLM with the Mamba model, and the proposed Stitch-Scan is merely a data augmentation stitching method.\n\n- The experiments are insufficient. The core argument of this article is: \"we first construct visual long sequences with multiple vision encoders, which not only enrich visual representations but also leverage the advantages of Mamba in handling long sequences. Notably, this design will not undermine the efficiency obviously, which is in stark contrast with the common cognition of Transformer-based MLLMs.\" Is there any experimental or theoretical support for this conclusion? How much is \"not undermining the efficiency obviously\" specifically? It is recommended that a row be added to Table 4 so that the visual tokens of MambaVLM and MobileLLaMA-2.7B are also consistent at 144, which would support the above point.\n\n- Formula 7 is expressed non-standardly; do not mix mathematical symbols with code.\n\n- In Formula 8, Hv = Merge(Hv1, Hv2, Hv3, Hv4), the Merge method is not explained in the text. What specific merging technique is used, just a simple concatenation?\n\n- In Table 1, the Qwen-VL model outperforms MambaVLM in performance on TextVQA and VQAv2 with a data scale of 665K. Typically in papers, bold numbers indicate the best results obtained by models, but this is not the case in your table. If the bold numbers have a special meaning, please explain this in the text. Additionally, the same issue occurs in Table 2." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "See Weakness" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The framework is concise and clear, making the proposed approach easy to understand." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a customized version of the Mamba framework within multimodal large language models (MLLMs). This framework has three core components: a visual long sequence, a Mamba projector, and a Mamba LLM. Experimental results on various benchmarks suggest improved performance and speed compared to several existing methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1): Limited Novelty in Technical Contribution: While the paper proposes a \"visual long sequence\" as part of the framework, a significant body of literature already exists on augmenting visual features using ensembles of different visual encoders, as demonstrated in works such as [A-D]. The design of the Mamba projector, specifically its cross-stitch scanning scheme that concatenates four scanning paths, seems heuristic rather than theoretically grounded.\n\n2): Unclear Motivation for Mamba Projector: The Mamba projector, the primary technical contribution of this paper, has an unclear motivation. The 1x1 convolutional MLP layer can be treated as a full attention layer, suggesting that the Mamba projector is an approximation. Lines 250–253 argue that \"a simple MLP layer may not be able to accomplish sufficient vision-language alignment and interaction of different visual features. Therefore, we devise a lightweight mamba projector…\" However, this rationale does not sufficiently justify the addition of the Mamba projector.\n\n3): Unfair Experimental Comparisons: For instance, in Table 4, using a longer visual sequence generally increases latency. Models such as TinyLLaVA and MobileVLMv2 should be substituted with the Mamba LLM. In Table 2, MambaVLM shows superior performance, largely attributed to encoder ensembling—a common approach in the literature.\n\n4): Presentation Quality: The paper’s overall clarity and presentation could benefit from further refinement.\n\n\n\nReferences:\n\n[A]: BRAVE: Broadening the Visual Encoding of Vision-Language Models, ArXiv.\n\n[B]: Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs, CVPR 2024.\n\n[C]: Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders, ArXiv.\n\n[D]: Law of Vision Representation in MLLMs, ArXiv." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "See weaknesses." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The paper is easy to understand with clear illustrations on the proposed methods.\n2. According to the experiments, MambaVLM achieves overall better performance compared to previous VLMs, such as Qwen-VL, LLaVA-1.5, and Cobra." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a new variant of vision-language model (VLM) called MambaVLM, which introduces multiple improvements to previous VLM with Mamba method, Cobra. Specifically, the paper proposes to concat the visual features from DINOv2 and SigLIP by sequence axis instead of the channel-axis in Cobra, followed by a new Mamba-based projector. Performance is validated on various VLM benchmarks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The novelty is limited with the following reasons: (1) The model is based on Cobra, with minor changes on the concatenation of visual features and projector. (2) The scan directions are from VMamba, only the stitch-scan is novel.\n\n2. Although the sequence-level concatenation improves the performance, it poses a great concern on the efficiency of the model, but the authors did not provide the inference speed, computational cost, and memory cost comparisons. Though the Mamba has linear computational complexity, longer sequence indeed increases the FLOPs and memory consuption, and the heavy projector also introduces additional costs. As a result, directly compare the model with existing methods such as Cobra without comparing the efficiency is **unfair**.\n\n3. In Figure 1, directly comparing LLaVA-1.5 with MambaVLM to demonstrate the effective of Mamba and the superiority on training time is unfair, as MambaVLM uses better DINOv2-SigLIP encoder.\n\n4. In lines 215~235, \"regardless of how many channels ... loss of visual information\" is overstated, lacking precise theoretical evidence to support the claims. Bottleneck-structures are widely used in networks such as ResNet, and according to information bottleneck principle, it is no clear evidence to state that the compression of channels will definitely lose the valuable information. Please reword.\n\n5. In Table 1, some results (62.6, 76.3) of MambaVLM is not the best and should not be bolded. Please correct them." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see the section on weaknesses." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The proposed method in the paper performs very well, achieving better performance than LLaVA 1.5 with only half the training time.\n2. The approach is ingenious, using Mamba for long-context vision-language modeling is a promising avenue worth exploring.\n3. The paper is written with a clear structure." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "MambaVLM is a highly efficient multi-modal large language model framework that integrates Mamba’s linear complexity with a novel cross-stitch scanning approach to improve both visual information interaction and vision-language alignment. Achieving competitive benchmark results with only 0.66 million data points and 14 hours of training on a single A800 node, MambaVLM significantly outperforms LLaVA-1.5 and rivals the performance of Qwen-VL, demonstrating Mamba’s potential in enhancing MLLM efficiency and effectiveness." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The performance comparison with the original LLaVA is somewhat unfair, as the method in the paper uses two visual encoders. It would be better if a version with only ViT-CLIP could be provided.\n2. The method description in the paper is unclear; perhaps I missed where it explains how Mamba-VLM + Vicuna is implemented. It seems that if Vicuna is used, only the Mamba projector is related to Mamba. Of course, I also understand that the performance of VLMs is highly dependent on the performance of the LLM, and Mamba as an LLM is still relatively weak." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@misc{\nchen2024unleashing,\ntitle={Unleashing the Power of Selective State Space Models in Vision-Language Models},\nauthor={Honghao Chen and Yibing Song and Shoufa Chen and Chongjian GE and Kaiqi Huang},\nyear={2024},\nurl={https://openreview.net/forum?id=0A6f1b66pE}\n}" }, "abstract": { "value": "While emerging multi-modal large language models (MLLM) have demonstrated impressive advances, the quadratic complexity of their Transformer-based LLMs (3B or larger) inevitably leads to considerable computational overhead. On the other hand, the recently proposed selective state space model (i.e., Mamba) enjoys both model capacity and computational efficiency, making it an ideal component to enhance MLLM's efficiency and performance. However, recent attempts to introduce Mamba into MLLMs simply replace their LLMs with Mamba, ignoring the unique characteristics of either side. We argue that such a naive combination cannot exhibit the potential of Mamba in MLLMs. In this paper, we delve into harnessing Mamba's unique properties, and propose tailored designs from both multi-modal input and architectural perspectives to unleash its true power. First, we fully utilize Mamba's linear complexity to construct visual long sequences for a thorough perception at a minor efficiency burden. To integrate the scanning mechanism with the built visual long sequence, we devise a novel cross-stitch scanning approach to capture and fuse spatial and semantic properties simultaneously, enhancing the interaction of visual information and the vision-language alignment. Built upon these designs, we propose MambaVLM, a simple yet effective MLLM framework that exhibits highly competitive results across multiple benchmarks. Moreover, our framework is also compatible with Transformer-based LLMs (e.g., Vicuna), demonstrating remarkable training and inference efficiency. Notably, with only 0.66M data and 14 hours training on a single A800 node, our MambaVLM outperforms LLaVA-1.5 by significant margins and performs on par or even better than the 1.4B data trained Qwen-VL. The appealing results from both effectiveness and efficiency aspects indicate the promising prospects of Mamba in MLLMs." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": { "value": [ "~Honghao_Chen1", "~Yibing_Song1", "~Shoufa_Chen1", "~Chongjian_GE1", "~Kaiqi_Huang1" ] }, "authors": { "value": [ "Honghao Chen", "Yibing Song", "Shoufa Chen", "Chongjian GE", "Kaiqi Huang" ] }, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Vision-Language Models; Mamba;" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": { "value": "chen|unleashing_the_power_of_selective_state_space_models_in_visionlanguage_models" }, "pdf": { "value": "/pdf/1183a88cc6949631a83ff61a73451d323b4592a6.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Unleashing the Power of Selective State Space Models in Vision-Language Models" }, "venue": { "value": "ICLR 2025 Conference Withdrawn Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Withdrawn_Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 2 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "No ethics concerns" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. What are some other regularity verification methods? how does the SOS programming compare to them in terms of accuracy and efficiency?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper’s strength lies in its innovative application of Sum-of-Squares (SOS) programming to address the longstanding challenge of verifying the regularity of optimal transport (OT) maps through the Ma-Trudinger-Wang (MTW) tensor. SOS programming is a well-established tool in optimization and control, but this work extends it to OT regularity, opening new possibilities for computational verification of the MTW conditions in general cases where analytic approaches are intractable. The paper also showcases the practical efficacy of the approach by applying it to various cost functions, demonstrating its flexibility and adaptability across different scenarios." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a computational approach using Sum-of-Squares (SOS) programming to verify and approximate regions of regularity for optimal transport maps, specifically focusing on the Ma-Trudinger-Wang (MTW) tensor. Regularity of optimal transport maps is critical in many machine learning applications, and this regularity can be assured by the non-negativity of the MTW tensor. However, verifying this condition analytically for general cost functions is challenging. The authors propose using SOS programming to generate certificates of non-negativity for the MTW tensor across a broader range of cost functions, potentially providing computationally verified regions of regularity. Their method is applied to both verifying non-negativity conditions (the \"forward problem\") and to computing inner approximations of regularity regions (the \"inverse problem\") for several ground cost functions, demonstrating the effectiveness of SOS programming in this context. This computational framework contributes a systematic approach to certifying regularity in optimal transport, potentially facilitating its application in various machine learning tasks. The paper concludes by applying the proposed framework to several common examples in the literature." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "A missing key aspect in the paper is the time complexity analysis for the proposed framework. What's the computational efficiency of SOS programming in verifying regularity of the different OT problems? While the authors showcase the method’s application to specific examples and shared the wall-clock time, a time complexity discussion could be a good addition to the paper." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "My main question pertains to how the authors see this work fitting within the broader optimal transport literature. In effect, though some results require regularity of the optimal transport map to hold, these results typically pertain to questions of statistical estimation. In these settings, population measures are estimated based on samples and so (i) absolute continuity cannot be verified a priori, (ii) upper and lower bounds for the density cannot be verified, and (iii) the support of the distributions are unknown. It is thus unclear to me how the content of the current paper fits within the previous context." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "To my knowledge, this is the first paper which explores the question of numerically verifying the MTW condition. The paper is overall written well, and the theoretical details look correct." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper presents a sum-of-squares programming based approach to verifying the Ma-Trudinger-Wang (MTW) condition. Precisely, both the forward problem of identifying if a given cost function and domains satisfy the MTW condition and the inverse problem of finding the largest semialgebraic domain on which the MTW condition holds are considered. The corresponding problems can be solved via standard SOS solvers on modest hardware. The paper concludes with a numerical study which validates the theoretical findings." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The limitation of the paper regards the assumption that the cost function is rational or that the elements of the MTW tensor are rational. I believe it would be useful to provide general examples of when it holds/does not hold in the text to further clarify how strong/weak the assumption really is. \n\nI believe, however, that the implications of this work are not quite fully fleshed out." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 2 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Could you please provide examples of real-world OT applications where the knowledge of the regularity of the Monge map is crucial ?\n- Could you please bring more details on the equivalence between non-negativity on polynomial terms and SOS representation ?\n- I think it would be of interest to provide the results on the regularity of the Monge transport map given the three types of non-negativity conditions given in the paper.\n- Have you considered experiments with higher dimension ? Is there any computational burden ?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- Although I am no expert of SOS programming, the paper is well written so that it can be read by a large audience. In particular, the notation is easy to understand and Section 2 provides the most essential theoretical elements from OT and SOS programming domains to introduce the method.\n- Given the elements of Section 2, the idea of proving the MTW non negativity via SOS programming seems to be a very good (and natural) idea. This work seems to be the first to answer this question with relatively moderate theoretical and computational frameworks.\n- The formulation of the inverse problem is very interesting and once again well introduced and explained.\n- The diversity of numerical experiments (i.e. non trivial transport costs) definitely proves the theoretical statements." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In this paper, the authors propose the first computational approach to certify the regularity of the Monge map for optimal transport problems with specific conditions on the transport cost and the state spaces. To be more precise, they evaluate the non-negativity of the fourth-order Ma-Trudinger-Wang (MTW) tensor associated to the transport cost, which has been proved to be a sufficient condition to establish the continuity of the Monge transport map under proper conditions on the marginals of the transport plan. In this work, they consider three versions of this non-negativity condition, previously considered by related works. Their method consists in reformulating the MTW condition (for each of the three versions) into a sum-of-squares program defined on a semialgebraic setvia Putinar's Positivstellansatz, which can then be solved with efficient software. In particular, their approach assumes that the transport cost (or at least the corresponding MTW tensor) is a rational function defined over a two-state semialgebraic space (or at least, a two-state space that contains a semialgebraic space). They apply their framework to verify if the transport cost verifies the MTW condition or to find the largest semialgebraic set on which the transport cost verifies the MTW condition. They propose several convincing numeric experiments in small dimension for a large variety of non-trivial transport costs." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- As it seems crucial to apply SOS programming, the transformation of the non-negativity condition into a SOS representation in Equation (8) would deserve more explanation, in the appendix for example. For non expert readers, this relation is hard to understand.\n- The dimension of the numerical experiments is relatively low, while OT aims at solving large-scale problems.\n- Although the problem tackled in this paper is interesting from a theoretical perspective, I am quite concerned by the effective application of this work to OT problems." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 2 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "1. I encourage the authors to elaborate on potential ML applications or benefits of their work on non-Euclidean optimal transport. For example, would you like to discuss how your method could enhance practical ML systems that use OT, or to provide concrete examples of where non-Euclidean costs arise in ML problems? \n\n2. I encourage the authors to improve accessibility of technical parts (e.g., the description of forward problems and inverse problems), such as adding more intuitive explanations or examples alongside the formal mathematical notation." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "This is a mathematically solid paper and resolve an interesting theoretical problem. It proposes a provably correct computational framework that can certify or falsify the non-negativity of the MTW tensor associated with a given ground cost under the assumptions that the ground cost is a rational and semialgebraic function. The proposed approach is based on sum-of-squares (SOS) programming and can be of independent interests. The authors also demonstrate that the proposed computational framework can be applied to non-rational ground cost function given that the elements of the MTW tensor are rational and can be used to solve the inverse problem." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In the context of OT, the fourth-order Ma-TrudingerWang (MTW) tensor associated with this ground cost function provides a notion\nof curvature. The non-negativity of this tensor plays a crucial role for establishing continuity for the Monge optimal transport map. In general, it is difficult to analytically verify this condition for any given ground cost. This paper proposes a provably correct computational approach which provides certificates of non-negativity for the MTW tensor using Sum-of-Squares (SOS) programming. The authors further show that their SOS technique can also be used to compute an inner approximation of the region where MTW non-negativity holds. They apply this proposed SOS programming method to several practical ground cost functions to approximate the regions of regularity of the corresponding OT maps. They also evaluate the proposed SOS computational framework for both the forward and the inverse problems." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The main concern is that this paper seems irrelevant to the ICLR community. In practice, the common ground cost function would be Euclidean and I am not really sure if it is practically important to conduct the computational verification of OT regularity for a general class of non-Euclidean ground cost functions. I encourage the authors to elaborate on potential ML applications or benefits of their work on non-Euclidean optimal transport. For example, would you like to discuss how your method could enhance practical ML systems that use OT, or to provide concrete examples of where non-Euclidean costs arise in ML problems? \n\nAnother concern is the poor quality of writing. In particular, there are many advanced mathematical notations, such as semialgebraic functions and Archimedean sets, which are not accessible to the ICLR audience. Both Section 2 and Section 3 are written in a technical way without sufficient intuitive explanations or examples alongside the formal mathematical notation. In my humble opinion, the major contribution of the paper would be the SOS formulations for computing the MTW tensors, which is certainly nontrivial, but this paper would much better fit the applied mathematics oriented journal." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The manuscript doesn't contain (A) a discussion on the computational complexity for checking the NNCC condition and the two types of MTW conditions. Also, the discussion on (B) the complexity of the inverse problem is missing. For (A), the author is advised to provide an analysis. For (B), the author is advised to also provide a runtime in the work, i.e. the time it takes to plot the figures.\n\n2. What is the relationship between Monge OT map (which is Borel) and a Brenier map (which is point to point)? Is regularity of Brenier map, supposing it exists, also something that this formulation can answer?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper presents an under-studied aspect of regularity of Monte map. Moreover, the paper contains numerical algorithm and is practical.\n2. The inverse problem is well thought through. \n3. The numerics seems promising for a relatively difficult problem in OT." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a computational method for assessing the regularity structure of optimal transport plans. Specifically, the source and target distribution are continuous distributions on a manifold, the cost is some smooth function, and the regularity in question is the regularity of the pushforward map from source to target distributions. In addition, the computational tool computes the region where the MTW holds." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. While it is not an issue for the author per se, it is unfortunate that the SOS condition only applies to semialgebraic sets for the manifold. \n\n2. The writing doesn't seem to include sufficient focus on the gap between the non-negativity condition and the SOS counterpart. It doesn't seem to be clear whether SOS is too strong for this case. \n\n3. The presentation is not clear for this paper. For example, the indicting convention for something like $c_{ij, p}$ is quite confusing. The only mention of $c_{ij, kl}$ in the earlier part is too far away, and the readers can not be expected to find where the notation is and also generalize from $c_{ij, kl}$ to $c_{ij, p}$. This is too confusing for this conference.\n\n4. Theorem 5 seems currently wrong: the function $F$ in (5) is matrix-valued. It doesn't seem correct to somehow assess if this matrix-valued rational function belongs to \\sum_{SOS}[x, y]. *Unless this issue is either resolved or explained, this reviewer cannot increase the score above the acceptance threshold.*\n\n5. The author doesn't seem to use certain terms in differentiable geometry correctly. For x, y on different points of $\\mathcal{M}$, one cannot directly apply an \"inner product\"/contraction between the tangent plane on $x$ and the contingent plane of $y$. This issue is resolved, however, if $\\mathcal{M}$ is a subset of $\\R^{n'}$ and the differentiable structure comes from the Euclidean space. The author is advised to change the writing on this and make sure no further major mistake such as this is made.\n\nMinor comments:\n- The logic at line 293 is wrong: \\eta(\\xi) = 0 should belong to after \\forall." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose a sum-of-squares programming method to numerically certify Ma-Trudinger-Wang regularity of optimal transport maps" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024sumofsquares,\ntitle={Sum-of-Squares Programming for Ma-Trudinger-Wang Regularity of Optimal Transport Maps},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0AHkdAtFW8},\nnote={under review}\n}" }, "abstract": { "value": "For a given ground cost, approximating the Monge optimal transport map that pushes forward a given probability measure onto another has become a staple in several modern machine learning algorithms. The fourth-order Ma-Trudinger-Wang (MTW) tensor associated with this ground cost function provides a notion of curvature in optimal transport. The non-negativity of this tensor plays a crucial role for establishing continuity for the Monge optimal transport map. It is, however, generally difficult to analytically verify this condition for any given ground cost. To expand the class of cost functions for which MTW non-negativity can be verified, we propose a provably correct computational approach which provides certificates of non-negativity for the MTW tensor using Sum-of-Squares (SOS) programming. We further show that our SOS technique can also be used to compute an inner approximation of the region where MTW non-negativity holds. We apply our proposed SOS programming method to several practical ground cost functions to approximate the regions of regularity of their corresponding optimal transport maps." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Optimal transport", "sum-of-squares programming", "Ma-Trudinger-Wang tensor" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/cab80dae9587724cfe6adbe63d723502452d5177.pdf" }, "presentation": null, "primary_area": { "value": "optimization" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/03d9034187d55d4efeb0b054e88695a7d9b9fd88.zip" }, "title": { "value": "Sum-of-Squares Programming for Ma-Trudinger-Wang Regularity of Optimal Transport Maps" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "See above." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper has a natural motivation, and the concluded rate seems matching that of classical Euclidean case. The presentation is lucid, and the proof sketch and extended discussion is well written. Overall this paper is a solid contribution on the topic of manifold learning." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper studies the complexity of Sobolev function class on Riemannian manifolds. Specifically, the paper derives lower bound of the approximation error of a Sobolev ball by a smaller class with complexity bounded as pseudo-dimension. By constructing explicitly functions of bounded Sobolev norm that are separated in $L^1$, the paper connects the packing number of the manifold with a hard-to-learn subclass in the Sobolev ball, thus forcing a larger error/width. The main theorem claims a lower bound that only depends on intrinsic quantities of the manifold." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The result is not too surprising on the 1,p-Sobolev class, and considering that higher Sobolev space can even be an RKHS [1], one would expect major improvement on the rate. Also using volume comparison to control the packing number is rather standard, and one might further ask if the same technique is applicable to metric measure spaces or RCD spaces, though I understand this technicality may not be particularly befitting of this venue.\n\n\n2. Typos: \nDefinition 2.7 is defining packing number not covering number, and also metric entropy is not commonly defined this way. This version of metric entropy is exactly the same as packing number, hence (7) is not needed. (The key proposition C.1 is correct.) $CPf_a$ out side the balls on line 375 is not necessarily 0.\n\n[1] De Vito, Ernesto, Nicole Mücke, and Lorenzo Rosasco. \"Reproducing kernel Hilbert spaces on manifolds: Sobolev and diffusion spaces.\" Analysis and Applications 19.03 (2021): 363-396." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- The Theorem 1 lower bound (12) has a dependence on $p$, unlike in the Euclidean case (Maiorov and Ratsaby, 1999). Is there a plausible reason for this, i.e. is $p$ there for an inherent reason? \n\n- Why is Theorem 1 restricted to the case $K < 0$? What is different about the positive curvature case?\n\n- line 535: it is mentioned that the cutoff functions with bounds on higher derivatives is difficult to construct, but I am having trouble seeing why this should be so. Can the authors explain further?\n\n- line 758: The authors say \"By maximality, balls of radius $2\\epsilon$ at the $p_i$ cover $M$\" but why is this true? The manifold can potentially be very narrow.\n\n- line 691: What is the notation $\\mathcal{A}(z, 16 / \\sqrt{\\text{length}(z)})$?\n\n- Should the empirical risk in (56) be scaled by the sample size?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "For a technical paper, the presentation is approachable and is self-contained (modulo some typos and missing definitions). The paper extends the lower bound proved in (Maiorov and Ratsaby, 1999) to manifolds." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper concerns a lower bound for a nonlinear width (involving the pseudo-dimension, a generalized version of the VC dimension) of Sobolev classes over smooth manifolds of dimension $d$. The authors claim that while the manifold can be embedded in a higher dimensional space with dimesion $D \\gg d$ the width of the Sobolev class has a lower bound that depends only on the dimension $d$ of the manifold." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Not only is the result of Theorem 1 independent of ambient dimension $D$, the ambient dimension does not appear _anywhere_ in the estimates. This is somewhat odd because the abstract mentions the manifold hypothesis which concerns both $D$ and $d$. In some similar approximation bounds, there is typically some dependence on $D$. The authors should address this.\n\n- In a similar vein, the authors do not present a connection between the sample complexity mentioned in Proposition 2.3 and the main Theorem 1, as far as I can see. The assumed connection is that, due to this property of classes with finite pseudo-dimension $\\mathcal{H}_n$, the Sobolev class can also be estimated with the sample complexity given in (2), once the approximating class $\\mathcal{H}_n$ is determined, it can be estimated with this sample complexity. This connection should be made somewhere.\n\n- The main structure of the proof is almost identical to (Maiorov and Ratsaby, 1999), except the construction of the $L^1$-separated set of functions, due to the domain being a manifold. There are a few questions about the extended lower bound, which I ask below in the \"questions\" section." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "I don't understand why the main result, Theorem 3.1, is measuring nonlinear n-width between the Sobolev space W^{1,p} and the Lebesgue space L^q. It's really unclear to me why this implies anything about the difficulty in approximating Sobolev space functions. I'd like to see this clarified." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "The question of why deep neural networks work well on extremely high dimensional data is an important problem and the manifold hypothesis may be a good way to explain this. Work on this problem is important and valuable in machine learning. The problem of lower bounds on complexity is not studied as often as upper bounds. The results appear to be new and non-trivial." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper is focused on the manifold assumption in machine learning. The goal is to further shed light on how the intrinsic dimension of the data manifold enters into notions of complexity in function approximation. In particular, the authors prove lower bounds on the complexity of approximating Sobolev space functions on manifolds, and show that the lower bounds, which are essentially 1/n^(1/d), depend only on the intrinsic dimension d of the manifold, and not the ambient dimension of the space the manifold lies in. The authors use a notion of pseudodimension that is an extension of VC-dimension and measure complexity by the nonlinear n-width." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "It's not clear to me how applicable these results are in practice. Even when data satisfies the manifold assumption, the intrinsic dimension d may be quite large. It is not clear how large the authors think d is in practice, and how large a d would make these results applicable vs vacuous. For MNIST, for example, it's often given that d is between 7 and 14, depending on the digit. One can assume d is much larger for more challenging problems, maybe 20-40? In this case, the error bound 1/n^(1/d) is vacuous, unless the number of data points n is astronomically large (e.g., if d=20 we need 10^(20) data points!)." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "For Riemannian manifolds, we provide lower bounds on approximating bounded Sobolev balls with classes of finite statistical complexity. The derived rate at which the lower bound converges to zero depends only on intrinsic properties of the manifold." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024blessing,\ntitle={Blessing of Dimensionality for Approximating Sobolev Classes on Manifolds},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0ASCZrVzSX},\nnote={under review}\n}" }, "abstract": { "value": "The manifold hypothesis says that natural high-dimensional data lie on or around a low-dimensional manifold. The recent success of statistical and learning-based methods in very high dimensions empirically supports this hypothesis, suggesting that typical worst-case analysis does not provide practical guarantees. A natural step for analysis is thus to assume the manifold hypothesis and derive bounds that are independent of any ambient dimensions that the data may be embedded in. Theoretical implications in this direction have recently been explored in terms of generalization of ReLU networks and convergence of Langevin methods. In this work, we consider optimal uniform approximations with functions of finite statistical complexity. While upper bounds on uniform approximation exist in the literature in terms of ReLU network approximation, we consider the opposite: lower bounds to quantify the fundamental difficulty of approximation on manifolds. In particular, we demonstrate that the statistical complexity required to approximate a class of bounded Sobolev functions on a compact manifold is bounded from below, and moreover that this bound is dependent only on the intrinsic properties of the manifold, such as curvature, volume, and injectivity radius." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "approximation theory", "manifold hypothesis", "statistical complexity", "Riemannian geometry" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/63ca09a40742fe6d4c6f669525b90ff9f172509d.pdf" }, "presentation": null, "primary_area": { "value": "learning theory" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Blessing of Dimensionality for Approximating Sobolev Classes on Manifolds" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Could the authors provide clarification on the points I raised in the weaknesses section, especially if I may have misunderstood some of the contributions?\n\n2. In Figure 6, do the authors have any insights into the concave behavior of the scaling factor? Are there specific explanations or potential methods for identifying this optimal scaling factor?\n\n3. Regarding the stop word shift in distribution, is it generally accepted that a higher probability of stop words negatively impacts LLM performance?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper is well-written and effectively illustrates the importance of superweights and superactivations. I appreciate the discussion on the percolation of superactivations across the network and the identification of superweights across layers (Figure 3). Additionally, I find the potential implications of superweight upscaling presented in Figure 6 quite interesting." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper focuses on the impact of outlier weights in large language models (LLMs), specifically larger weights, which the authors term superweights and superactivations. First, the authors analyze how much these weights and activations affect LLM performance. They then use this as motivation to discuss quantization methods designed to account for superweights and superactivations. Throughout the paper, the authors also discuss the impact of superweight scaling and provide experimental results showing how their quantization method improves upon standard rounding, especially when using larger block sizes within the network." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While I appreciate the analysis presented in this paper, I am struggling to see the novelty of this work. I may be misunderstanding, but from what I gather, superweights and superactivations have already been discussed in prior analyses of LLMs. Additionally, it seems that methods like AWQ and SqueezeLLM inherently focus on superactivations. Furthermore, compared to other weight quantization techniques, the proposed method does not appear to offer significant improvements." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "The paper mostly focuses on post training model weight/activation analysis and identifies certain handful of importance weights/activations. The authors also say that irrespective of the input prompt the super weights are always the same and they mostly occur in the early layer's down projection with some reasoning via skip connections diagram. \n\nThough these insights are helpful, but it would be good if authors can follow up with what happens during the training process that such super weights are formed in the first place. Does the training methodology in terms of quantization during training/layernorm, gradient scaling, etc play any role in the forming of these super weights?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "Novel discovery about the importance of a few handful of neurons: The identification and analysis of super weights and super activations as critical outliers and their positive influence on model's performance is noteworthy and interesting. \n\nQuantization proposals: Authors went one step further to propose a super weight-aware quantization method to make the best use of these super weights/activations. Data free quantization proposal with on par performance compared to SmoothQuant is also a worthy contribution." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper is about the discovery of super weights in LLMs that are disproportionately important, pruning these hurts model quality quite a bit. The authors have provided a way to identify these super weights using a forward pass. Super weights are activations are sensitive to quantization effects and hence authors propose a super weight aware quantization method enabling effective quantization." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Though the discovery is quite interesting, the improvements of proposed methods with existing baselines are quite marginal. In general, such kind of super weights might be a natural phenomenon in any machine learning model. How can one say this is relevant only to LLM's?\n\nThe work seems to be very much based on empirical observations (which is not my concern) but more discussions/intuitions/explanations around how/why these super weights are formed will be useful." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 1 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "Please refer to the weaknesses section above." }, "rating": { "value": 1 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "The authors conducted experimental explorations on the so-called \"super weights.\"" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper investigates the sensitivity of a subset of outliers in LLMs, referring to them as \"super weights.\" The authors conducted experiments to examine the impact of these super weights on model performance." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The necessity of \"super weights\" is unclear, as outliers are already identified based on the threshold. Increasing the threshold will naturally reduce the number of outliers with very large weights. Given the known importance of outliers in LLMs, emphasizing \"super weights\" (outliers at a higher threshold) does not appear novel.\n\n2. Figure 1 is misleading. According to the author's definition, \"super weights\" are a subset of outliers. However, the figure suggests -1.9 is a typical outlier with nearby values being quite small (.1 and .2), implying that zeroing out outliers produces nonsensical text—a widely acknowledged fact. To better demonstrate the significance of super weights, it would be beneficial to explore whether zeroing out all outliers results in poor performance, and similarly, whether zeroing out just a small subset (e.g., 20-30) leads to comparably severe degradation.\n\n3. Table 1 raises critical concerns. First, the criterion for selecting outliers needs specification. Second, the \"Prune SW, +SA\" setting in Lines 146-152 is confusing, as it suggests pruning super weights while partially restoring super activations enhances quality. However, the authors did not prune activations, leading to confusion about this claim.\n\n4. Table 2 appears redundant and fails to convey meaningful information. Replacing it with visual representations of \"super weights\" distributions would be more informative, as the current table occupies considerable space without offering clear insights.\n\n5. Figure 2 is difficult to interpret. The depiction of super weights and their impact, such as generating nonsensical text, is not clear. The use of the same color block in both the network and the output is puzzling. Are the model's dynamics linear? How do the output and weights share the same significance? Clarification is needed on whether this figure is based on assumptions or empirical data.\n\n6. In Lines 189-190, the term \"super activations\" is introduced but lacks clarity on whether it is threshold-based or aligns with corresponding weights, which could be time-consuming. The authors should clarify this terminology.\n\n7. The paper contains several unprofessional notations. For example, \"Yij\" should be corrected to \"Y_{ij}\" in Line 204, and similarly, \"Xik\" and \"Wjk\" should be \"X_{ik}\" and \"W_{jk}\" in Line 205. The inconsistency in notation and dimensions between \"d\" and \"D\" in Line 204 suggests a lack of careful writing and review, raising concerns about the overall professionalism of the paper.\n\n8. Lines 198-210, which discuss the identification of super weights, are crucial yet unclear. The selection criteria for super weights remain ambiguous and need a precise mathematical description. Readers should understand the definition of outliers and the criteria for their selection explicitly.\n\n9. The paper lacks consistency in terminology. \"Super weights\" sometimes refer to both activations and weights, and at other times only to weights, adding confusion. In Line 306, the term \"super outliers\" is introduced, suggesting that the paper should maintain consistent terminology from the start, including in the title, if both weights and activations are discussed.\n\nAfter several careful readings, there are numerous additional concerns throughout the paper. The issues are substantial and critical, making it unlikely to meet the standards of ICLR. I recommend a strong reject based on the quality of this paper and will not change my rate." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. For equation 1, the median is used to replace super activation. Is getting the median\ntime-consuming since GPU is not good at sorting? (Although there are GPU-version\nsorting algorithms)\n2. The authors mentioned that SmoothQuant does not report on some models this paper\nevaluates, they compare our results with naive W8A8 quantization (line 407 - line 409).\nCan the authors run SmoothQuant on these methods since it is open-source? The naive\nW8A8 is a too-weak baseline." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The discovery is interesting and the proposed quantization method is easy to implement, which\ncan maintain better performance compared to Round to nearest quantization with the same\nblock size." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper reveals that Large Language Models (LLMs) contain a very small subset of weights\n(super weights) that are extremely important, where removing them severely degrades model\nperformance. The researchers developed an efficient, data-free method to identify these super\nweights using only a single forward pass. They further investigated how these super weights\ninfluence network behavior by analyzing their relationship with activation outliers. Building on\nthese insights, they proposed a quantization approach that carefully preserves these super\nweights while effectively compressing other weights, resulting in the maintenance of model\nquality after compression." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The authors failed to show how the proposed methods can improve the SOTA.\n1. Although the method is data-free, its performance does not exceed SOTA methods like\nSmoothQuant, given incorporating a small calibration dataset would not increase the\nquantization complexity much.\n2. The author mentions that this method is hardware-friendly, but no experiments to show\nits effectiveness in improving latency, throughput, memory usage, etc." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Section 3.2, \"Prune SW+SA\": The description of the \"Prune SW+SA\" condition in Section 3.2 is unclear. Specifically, how does this condition differ from the original model? I understand that super activations typically precede super weights in the model. Therefore, I am unsure what modification is being made in \"Prune SW+SA\" and how it distinguishes itself from the original, unpruned model. Could you please elaborate on this procedure?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The identification of \"super weights\" and their connection to super activations represents a novel and potentially significant finding in understanding the inner workings of LLMs.\n- Connection of \"super weights\" to quantization accuracy is quite interesting and has practical implications.\n- The paper provides a clear methodology for identifying super weights and evaluating their impact, along with an index of super weight coordinates for common LLMs, facilitating further research." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces the concept of \"super weights\" in Large Language Models (LLMs), identifying a small number of individual weight parameters (as few as one) that have a disproportionately large impact on model performance. Pruning these super weights drastically reduces the quality of generated text, while pruning thousands of other larger-magnitude outliers has a negligible effect. The paper proposes a data-free method for identifying super weights based on their connection to \"super activations,\" exceptionally large activation outliers previously observed in LLMs. Finally, the paper demonstrates that preserving super weights and activations during quantization significantly improves compression quality, achieving competitive results methods like SmoothQuant." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "# Major\n\n- Connection to Adversarial Examples: The literature extensively documents how small changes in the input domain can drastically alter output probabilities. Consequently, significantly harming the network by removing weights, as demonstrated, is somewhat expected. A discussion addressing the connection between super weight removal and adversarial examples would strengthen the paper.\n\n- Magnitude Pruning Baseline: In Table 1, the comparison of super weight pruning with global magnitude pruning may not be the most informative. A stronger baseline would involve pruning only within the layer where super activations occur. This would better isolate the impact of the super weight itself.\n\n- Quantization Baseline: The \"Naive W8A8\" quantization baseline should incorporate clipping. The current presentation makes it unclear whether the observed improvements stem from outlier removal or clipping, especially since super weight handling affects only a single layer during quantization, while clipping is applied to every layer. Furthermore, it should be noted that the clipping threshold is determined using Wikitext-2, which is also included in the evaluation of quantized models.\n\n# Minor\n\n- Terminology: The term \"extreme\" might be more descriptive and informative than \"super\" when referring to these weights.\n\n- Weight Distribution Visualization: Including a histogram visualizing the position of the super weight within the overall weight distribution would enhance understanding of its magnitude relative to other weights." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We discover and study \"super weights\" in LLM, which are very few in numbers yet crucial to model quality." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024the,\ntitle={The Super Weight in Large Language Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0Ag8FQ5Rr3},\nnote={under review}\n}" }, "abstract": { "value": "Recent works have shown a surprising result: a small fraction of Large Language Model (LLM) parameter outliers are disproportionately important to the quality of the model. LLMs contain billions of parameters, so these small fractions, such as 0.01%, translate to hundreds of thousands of parameters. In this work, we present an even more surprising finding: pruning as few as a single parameter can destroy an LLM’s ability to generate text—resulting in an increase in perplexity by three orders of magnitude and reducing zero-shot accuracy to guessing. We propose a data-free method for identifying such parameters, termed super weights, using a single forward pass through the model. Additionally, we find that these super weights induce correspondingly rare and large activation outliers, termed super activations. When preserved with high precision, super activations can enhance simple round-to-nearest quantization, making it competitive with state-of-the-art methods. For weight quantization, we similarly find that by preserving the super weight and clipping other weight outliers, round-to-nearest quantization can scale to much larger block sizes than previously considered. To facilitate further research into super weights, we provide an index of super weight coordinates for common, openly available LLMs." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "natural language processing" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/42983b9a0531517e43e254e991be6f6a3f53eaf9.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "The Super Weight in Large Language Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please see my questions in the above weakness section." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The author critiques previous benchmarks for being either small in scale, limited to predicting chronological age, lacking standardized datasets, comparing only a limited number of models, or relying on mortality and disease data that have restricted access. \n\nThe proposed benchmark seems address all of these limitations. Derived from publicly accessible data, it includes processing of data from both age accelerating condition (ACC) and healthy control (HC) groups to test model’s ability to distinguish between these conditions. Diseases with ACC are well considered. The benchmark includes 4 well-defined tasks with a summary score and evaluates 13 previously published models." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The author introduces a benchmark designed to evaluate models of the epigenetic aging clock. The benchmark includes 66 datasets containing DNA methylation data that meet specific conditions and corresponding metadata, with a total sample size of 10,410. Four tasks are proposed to assess the models’ ability to distinguish between healthy individuals(HC) and age-accelerating conditions(ACC). Results of these four tests are summarized into Cumulative Benchmarking Score. The benchmark framework also includes 13 previously published models results." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The paper is well-written and comprehensive overall, but several technical points need further clarification:\n\n1. The selection of metrics for benchmark tasks requires more justification. Specifically, why do tasks 2, 3, and 4 report median instead of the mean? Additionally, task 4 mentions the \"presence of covariate shift,\" but this shift is not clearly explained. Could the authors specify the covariate shift further ?\n\n2. The rationale behind the summary benchmark score requires further explanation. Why was this scoring method chosen, and what are its advantages? Also, what does \"positive bias\" refer to in this context? In the Results section, it is stated that $S_{AA1}$ is adjusted by a ratio to penalize prediction bias, yet this concept of prediction bias remains unexplained. Further clarification on what prediction bias entails here would be beneficial.\n\n3. It appears that plots C and D in Figure 3 may be incorrectly presented. Plot D should likely represent $Med(|\\Delta|)$ rather than $Med(\\Delta)$, as all points are above the diagonal. Please clarify if this is a mislabeling or if I have misunderstood the data shown." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "- Will you make your benchmarking dataset publicly available? Can you please add a link to it in your manuscript? I view this benchmarking dataset as a significant portion of your contribution in this work.\n- Can you please confirm that your evaluation tasks/metrics are original, and add citations if not?\n- Can you make a case for why the paper is a strong fit for ICLR, despite not truly being in the representation learning space?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- This paper is written very clearly, and did a great job walking the reader through the background to the problem, definitions of biological age, and different kinds of biological clock models. It’s graphics are informative, clear, and aesthetic. Truly a pleasure to read! \n- Provides colab notebook for reproducibility\n- I believe this paper will be significant to those in the biological clocks community. It is a benchmarking paper, so while it doesn't offer a new methodology itself, it does offer original tasks/metrics for assessing the performance of these models (I think they are original, I asked for clarification in the questions section) and a standardized benchmarking dataset (I asked for clarity to confirm it will in fact be published along with this paper)" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper benchmarks 13 different published biological clock models using a standardized test dataset that they compiled from more than 50 different publicly available studies. While no ground truth data is available for biological age (as it is a latent factor) or for age at death (as this data often isn’t published), the authors offer 4 compelling metrics by which to score the models accuracy and robustness. This paper presents a resource to the community in terms of a newly published benchmarking dataset, well-motivated metrics, and ratings for the current state of the art clock models. The paper also appropriately outlines limitations, such as the fact that some datasets had poor performance across all models, raising questions about dataset shift and for what kinds of data the clocks can be expected to make sound predictions. I believe this paper will help generate scientific discussion and progress in the aging clocks research community." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- I was disappointed that the clock models weren't all re-trained on a standardized training dataset. Without standardizing the training data, it is impossible to know whether the methodology of the clock or the training data it used are contributing to better/worse performance. This insight would be critical to the community in improving clock methodologies going forward.\n- The way that the authors chose to combine benchmarks in the cumulative score requires more justification. I am not sure why the different metrics should affect each other's weights so much. A simple sum, or weighted sum, of the four variables might be more appropriate if stronger justification is not supplied.\n- Requires clarification: on the one hand, authors write \"Clearly, the first task [AA1] provides a more rigorous way to test aging clocks [compared to AA2]\" on the other hand, they write \"The most rigorous of the four, AA2 task demonstrates...\"\n- Your description of the biomarker paradox could be improved. When I first read your description, I was left with questions. I had trouble finding more info on the \"paradox of biomarkers\" using the papers you cited (possibly due to paywall issues, I couldn't see the full articles), but you might consider adding this reference _Sluiskes, Marije H., et al. \"Clarifying the biological and statistical assumptions of cross-sectional biological age predictors: an elaborate illustration using synthetic and real data.\" BMC Medical Research Methodology 24.1 (2024): 58._ as their explanation made me fully understand the problem, namely that \"a (bio)marker that perfectly correlates with chronological age is useless in estimating biological age... in principle a nearly perfect chronological age predictor can be developed, as long as the sample size is large enough [35]. In such a case all signal related to biological aging would be lost.\"\n\nMore broadly, while I really enjoyed the paper, I am not sure it is a great fit for the ICLR community, as this model is a predictive regression model and not in the space of representation learning." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "I would like to ask the authors to address the two main criticisms I listed in the \"weaknesses\" section:\n- Overall, the opinion of this reviewer is that while the work has undoubtedly merit, it would be better suited for a forum more specific to biological age and aging clocks. \n- Regarding the normalization of methylation data, I would invite the authors to at least discuss whether the preprocessing of the included datasets match the recommended preprocessing of each aging clock (if any)." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The benchmark is well structured: (i) a variety of datasets and methods are included, and (ii) the tasks upon which the methods are evaluated are well defined and relevant for the domain. Furthermore, such type of benchmarks are quite timely, due to a continuously growing list of available aging clocks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors present a benchmark study where they contrast different computational methods, namely aging clocks, for inferring biological age from epigenetics (methylation) data. A corpus of datasets relevant for the benchmark was built through a systematic search, and it is provided as a resource. Finally, the evaluation was performed on four different tasks, devised in such a way to capture different aspects of aging clocks' performances." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "My main criticism is that the paper is only marginally relevant with respect to the topics of the conference. Inferring the biological age of an individual can hardly be considered as learning representations. The machine learning methods used for deriving aging clocks are very well known and established, thus lacking novelty. The tasks presented in the paper to assess the clocks' performances are not totally novel, as the authors themselves point out in section 2.2.\n\nFrom a technical point of view, an important aspect that the paper does not address is preprocessing. Several normalization methods exist for methylation data, and their impact to downstream analysis is well documented (see for example Teschendorff et al. 2013). A robust benchmark should try to evaluate the effect of different normalization methods on aging clock performances. \n\nA minor issue the authors may want to consider: the long list of reference at page 6 could be placed in the appendix, to ease reading\n\nAndrew E. Teschendorff, Francesco Marabita, Matthias Lechner, Thomas Bartlett, Jesper Tegner, David Gomez-Cabrero, Stephan Beck, A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data, Bioinformatics, Volume 29, Issue 2, January 2013, Pages 189–196," }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "Yes, Responsible research practice (e.g., human subjects, data release)" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "### Questions for the Authors\n\nIn evaluating the dataset and methodology presented, several questions arose that could help clarify the dataset’s potential applications and limitations.\n\n1. **Applicability for Method Development**: Can this dataset be effectively used for developing new methods on epigenetic aging clocks, or is it primarily intended for benchmarking and evaluation? Are there features or structures in the dataset that support novel method exploration?\n\n2. **Data Diversity and Representativeness**: How does the dataset account for demographic and biological diversity? Could the authors provide more details on the inclusion criteria to ensure the dataset is representative of a broad population?\n\n3. **Addressing Balance and Bias**: Were any steps taken to balance the dataset across aging-accelerating conditions (AACs) and healthy controls, or to mitigate known biases in the sample selection process?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "### Strengths \n\n The paper is clear and well-written, providing a solid foundation for its contributions. It presents a unified framework for evaluating epigenetic aging clocks, covering both first- and second-generation clocks. By introducing a benchmark dataset, the authors enable comprehensive testing of multiple epigenetic clock methods. \n\nThis work has potential to significantly impact the field of biological aging, as it offers a standardized dataset that can facilitate consistent evaluation across various epigenetic clock methods. Such a resource will likely streamline method comparison and improve reliability in aging research." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes ComputAgeBench, a unified framework and benchmark dataset for evaluating epigenetic aging clocks, which are predictive models for estimating biological age based on DNA methylation data. The framework aggregates 66 public datasets covering 19 aging-accelerating conditions, along with 13 published epigenetic clock models, to assess model performance consistently across a standardized dataset. The methodology incorporates rigorous evaluation criteria to test each model’s ability to distinguish between healthy individuals and those with accelerated aging conditions." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "In reviewing the proposed benchmark in this paper, several key areas for improvement have emerged, particularly concerning data diversity, balance, and bias. \n\n \n\n### Weaknesses \n\n \n\n1. **Limited Report on Data Diversity**: The paper lacks adequate details on demographic and biological diversity, such as age, ethnicity, and health variations. Including these would improve the dataset's representativeness for broader applications. \n\n \n\n2. **Data Balance and Bias**: The authors do not address balance across categories (e.g., AACs vs. healthy controls) or potential sampling biases. This oversight may skew benchmarking results and limit generalizability. \n\n \n\n3. **Absence of Bias Mitigation**: No strategies are mentioned to detect or reduce dataset biases, which is crucial for fair benchmarking in aging prediction models, where demographic factors can affect DNA methylation patterns and model performance. Additional evaluation metrics for fairness would increase the strength of this benchmark. \n\n \n\n4. **Put Together Publicly Available Dataset**: The proposed dataset, to my understanding, is a collection of existing publicly available datasets. The authors do not present to the research community a new benchmarking dataset, they rather collect existing datasets that they put together with a published harmonization technique. \n\nThe fact that the datasets already exist publicly, reduces the novelty of the benchmark. However, I cannot ignore that putting together 66 datasets into a single dataset is a contribution that would facitilitate the comparison of epigenetic clock methods." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "ComputAgeBench is the first framework for benchmarking aging clocks, which comprises 66 open-access datasets and compares 13 published models to find reliable biomarkers of health and aging." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024computagebench,\ntitle={ComputAgeBench: Epigenetic Aging Clocks Benchmark},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0ApkwFlCxq},\nnote={under review}\n}" }, "abstract": { "value": "The success of clinical trials of longevity drugs relies heavily on identifying integrative health and aging biomarkers, such as biological age. Epigenetic aging clocks predict the biological age of an individual using their DNA methylation profiles, commonly retrieved from blood samples. However, there is no standardized methodology to validate and compare epigenetic clock models as yet. We propose ComputAgeBench, a unifying framework that comprises such a methodology and a dataset for comprehensive benchmarking of different clinically relevant aging clocks. Our methodology exploits the core idea that reliable aging clocks must be able to distinguish between healthy individuals and those with aging-accelerating conditions. Specifically, we collected and harmonized 66 public datasets of blood DNA methylation, covering 19 such conditions across different ages and tested 13 published clock models. We believe our work will bring the fields of aging biology and machine learning closer together for the research on reliable biomarkers of health and aging." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "biological age", "epigenetic aging clocks", "DNA methylation", "aging biomarkers", "longevity" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/9e19fd966cb870bf815b97603aafb24b14afab8b.pdf" }, "presentation": null, "primary_area": { "value": "datasets and benchmarks" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/7bee137257178917182aa27edd16b502970dafe1.zip" }, "title": { "value": "ComputAgeBench: Epigenetic Aging Clocks Benchmark" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. When the attributes of a dataset are not directly accessible, how can they be retrieved? \n\n2. The current method utilizes a GAN-based model as the foundation. Is it feasible to implement this approach using a diffusion model instead? \n\n3. Additionally, how many attributes can this method manage effectively? If we aim to train a general-purpose model that can handle more than a thousand attributes, what strategies should be employed to address this scenario?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- Compositional image generation is a critical and practical problem, and this paper proposes a method to address it. \n\n- The paper presents an identifiable guarantee for learning the underlying concepts. \n\n- The generated images are promising, demonstrating the potential of the proposed method." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper addresses an intriguing problem: compositional image generation. It introduces the minimal change principle and proposes a method to limit the information introduced by each label. A causal conditioning approach is employed to disentangle concepts from correlations. The effectiveness of this method is validated across several tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. This method relies on pre-defined attributes, which limits the method's practical applicability. \n\n2. Additionally, the proposed methods are evaluated only on simple datasets, which may not adequately represent complex real-world scenarios." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "* Regarding Section 3.4, it’s unclear why inversion cannot be done in the $z$ space or $w$ space. Would it be possible to move the input of $f\\_i$ to $z$ or $w$ space and perform inversion in $z$ instead?\n* It is unclear why the first row of Table 2 is labeled as \"Ours.\" It appears to correspond to StyleGAN2-ADA." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "* The minimal change principle is intuitive and makes sense.\n* The concept of causal conditioning is interesting and intuitive.\n* The proposed method achieves superior FID scores on MNIST4 and Car9 datasets compared to StyleADA and AugGAN." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents the minimal change principle and causal conditioning to allow generative models to create compositional images with clear, identifiable concepts. The central idea is to control image attributes without inducing unintended changes. To accomplish this, the authors regularize the model to learn the minimum dimensions needed to edit an attribute and use causal discovery algorithms to disentangle dependent attributes. The authors empirically and theoretically demonstrate that this approach enables models to learn attributes that are both identifiable and composable." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "* There is no quantitative comparison showing if the model controls attributes better than the baselines. For example, metrics like Editing \n FID from StyleRes could be used to demonstrate controllability.\n* Baselines about image editing and compositional image generation are missing.\n * CausalGAN (Kocaoglu, et al. \"Causalgan: Learning causal implicit generative models with adversarial training.\" 2017.)\n * AugGAN(on FFHQ) (Hou, et al. \"Augmentation-aware self-supervision for data-efficient GAN training.\" 2024.)\n * StyleRes (Pehlivan, et al. \"Styleres: Transforming the residuals for real image editing with stylegan.\" 2023.)\n * HyperStyle (Alaluf, et al. \"Hyperstyle: Stylegan inversion with hypernetworks for real image editing.\" 2022.)\n * StyleTransformer. (Hu, et al. \"Style transformer for image inversion and editing.\" 2022.)\n* Except for Figure 8, there is no metric provided for editability or composability, making it difficult to assess whether the proposed method learns more identifiable concepts than the baselines. Additionally, in the ablation studies, it is challenging to gauge the effectiveness of the proposed components without metrics for editability or composability." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- What is the role of $\\mathbf{z}_c^*$ in equation (1)? It seems like it should encode information not represented by annotated labels (e.g., nuanced details). However, isn’t this type of information typically handled by the random noise $\\epsilon$? Does including $\\mathbf{z}_c^*$ have a significant impact on performance?\n- What is the role of $\\mathbf{z}^{\\text{null}}_i$ in equation (6)? What kind of information is it intended to encode?\n- It is hard to fully understand why enforcing the sparsity loss in equation (7) induces the minimal change principle. While Lines 522–524 suggest that constraining the representation’s dimensionality limits redundant information, this rationale is not entirely convincing. The minimal change principle, as described by the authors, states that \"the influence brought by each ground-truth concept should be minimal,\" which implies that changes in representation space should translate to minimal changes in the output space (e.g., altering the ‘age’ should yield the same image but with a different age). However, the sparsity loss in Equation (7) seems to restrict the input representation space rather than the changes in the output space, making it unclear how this connects to the minimal change principle.\n- It would be better to use distinct notations for $\\mathbf{z}_i$ in equation (3) and (6) as they are clearly denoting different variables. \n- Does $\\mathbf{m}_i$ in L177 refer to $\\mathbf{A}_i$? \n- In Figure 6, the authors claim that foundation models (e.g., GPT-4o) generate unrealistic images for unseen attribute combinations. However, all images generated by GPT-4o in Figure 6 appear unnatural, suggesting that the poor results might not be due to rare attribute combinations but other factors, such as improper prompts provided to the model. Could the authors clarify if proper prompt was used, and whether different prompts might correct GPT-4’s performance on unseen combinations?\n- In Table 8, which evaluates generation performance on human faces, it would be more comprehensive to include metrics for other generative models (e.g., GPT-4o, Meta AI, Stable Diffusion 3, as in Figure 6) for comparison.\n- Between the sparsity condition and causal conditioning, which component is the key factor that causes the proposed method to succeed where the baselines fail in Figure 5? Would simply applying causal conditioning to the baselines improve their performance?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The idea of transforming labels to identify and disentangle causal relationships among attributes is interesting, and the authors have effectively demonstrated its impact in the experimental results.\n- The proposed method significantly outperforms the baselines, both qualitatively and quantitatively, validating its practical advantages in achieving high-quality, controllable image generation, even in low-data settings." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a GAN-based framework for learning identifiable concepts. Given ground-truth attribute labels, random noise is transformed into latent representations aligned with these labels, and sparsified using learnable masks to enforce a minimal change principle. To mitigate existing correlations between certain attributes, the authors explicitly identify causal relationships among attributes and factorize the labels to remove dependencies. Empirical results demonstrate that the proposed method outperforms baselines in terms of data efficiency and controllability.\nThe main contributions of the paper are as follows:\n- Formulation of the minimal change principle to learn compositional concepts, along with an efficient approach to factorize causally related attributes.\n- Theoretical proof that the proposed method can recover ground-truth concepts.\n- Empirical evidence showcasing improved data efficiency and controllability." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- It is unclear how the proposed method learns compositional concepts more effectively or in a fundamentally different way compared to existing approaches. Since the baselines also leverage disentangled ground-truth attribute labels, wouldn’t they similarly be capable of learning a generative model for compositional generation? In a similar context, it’s not fully explained why the proposed method is more data-efficient than the baselines. A more detailed elaboration on these points would strengthen the paper.\n- The paper introduces several components (e.g., sparsity loss, learnable masks, $\\mathbf{z}^{\\text{null}}_i$​), but the justification for each component and their connections seems weak. It is a bit confusing as a reader to understand why each part is necessary. Please refer to the questions below for specific points on this aspect." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "n/a" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "It seems a relatively big limitation that the method relies on such rigid one-hot labels when the modern paradigm of image editing involves free-form textual descriptions. Do the authors envision easy ways to extend this to continuous or multi-label attributes, or free-form text? A discussion of the proposed binary attribute paradigm relates to the common free-form text editing one -- and their relative strengths -- would be insightful here." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "- The studied problem of generating unseen attribute combinations is a pertinent one, with important implications for under-represented demographics and subpopulations. The authors did a convincing job with Figure 1 and in the introduction of motivating the problem with current large-scale image synthesis models, and making the benefits of the proposed solution salient.\n- I appreciate that the experiments are relatively thorough in exploring multiple forms of image synthesis. Not only do the authors consider unconditional synthesis, but they also show how one can edit real images, greatly improving the contribution of their method." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors study compositional attribute generation and editing in image synthesis models. They argue relatively convincingly that the current large-scale image generation models fail to generate uncommon attribute labels (e.g. “female” + “facial hair”), and propose a methodology to address this through the use of masks learned with a causal structure. The results show the method produces images that do not exhibit mode collapse like the baselines. In the case of editing real images, there is significant improvements to the editing of rare attribute combinations over recent work." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "## [W1] Trade-off in performance for common attributes\n\nIn Figure 7, whilst the method clearly excels at generating rare attribute combinations (e.g. female + goatee), it fails in other cases to make more common edits (e.g. +blonde hair, or +bald).\n\nTo me, this seems very problematic. Almost by definition, most users will be interested in generating common attribute combinations. The fact that the method works so well on unseen combinations is a testament to its potential value, to be clear, but trading-off functionality for common edits at the same time seems like a clear and fundamental limitation. What use case does the proposed method serve if it’s at the cost of the common attribute combinations? In my view, this is the primary issue with the paper.\n\nAt minimum, I would expect to see a detailed discussion of this trade-off, and a solid justification for why it is worth making. Do the authors have any insights into why this might be happening? Furthermore, an insightful study would be one that quantifies the \"accuracy\" of edits for common vs uncommon attributes -- one could train a CelebA binary classifier to classify if an edited image actually depicts the new attribute or not, and one could see a breakdown of the performance for common vs rare attributes.\n\n## [W2] Lack of convincing baselines for independent attribute datasets\n\nI am not convinced that the authors do a good job of showcasing the benefits of their method in the independent attribute setting (Table 1 and Figure 5). Concretely, it is worrying that the baseline methods mostly fail to generate anything coherent at all (~20x as large FID scores). This really does not tell us much other than the baselines failed to train well (which could be for any number of reasons).\n\nThe authors could do a better job training the baseline models for a fairer comparison (e.g. perhaps with significant data augmentation, or through differentiable techniques such as [1]). Ultimately, we are not interested in the image quality itself, but instead in how well they perform in the “Out-FID” row on the rare attribute combinations. Through better training of the base models, we can isolate the impact of the proposed method on this row of interest without the confounding variable of the raw image synthesis quality in the way.\n\n## minor\n\nThe paper is full of typos, and some poorly written sentences. Just to mention a handful of examples from the introduction alone on the second page:\n\n- [L64] leads to → lead to\n- [L66] Ssadow → Shadow\n- [L72] mkae→ make\n\nUltimately these typos are indicative of a lack of care for presentation, and at times this renders the sentences hard to parse which I found often detracting from the content of the paper. I suggest some careful proof-reading is needed before the camera-ready or resubmission.\n\n---\n\n[1] Zhao et al. “Differentiable Augmentation for Data-Efficient GAN Training.” NeurIPS 2020." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We present a method for compositional image generation and editing with identifiability guarantees." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024learning,\ntitle={Learning Identifiable Concepts for Compositional Image Generation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0BBzwpLVpm},\nnote={under review}\n}" }, "abstract": { "value": "Humans have the ability to decompose objects into parts and relationships and\ncreate new objects by properly combining existing concepts. However, enabling\nmachines to achieve this in real-world tasks remains a challenge. In this paper,\nwe investigate how to teach machines compositional image generation through\nlearning identifiable concepts. To derive concepts from attribute labels, we formulate the minimal change principle and propose a method to limit the information introduced by each label. Additionally, to address dependent attribute labels\n(with causal influences in between or common causes behind them), we present\na causal conditioning approach to disentangle concepts from these correlations.\nOur framework enhances data efficiency, interpretability, and control, while enabling sampling from unseen combinations. We validate our method on various\ncompositional image generation and editing tasks, demonstrating its effectiveness\nthrough superior performance." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "concept; composition; image generation" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/1746275f4e677f128e95eb27c5e373f2d90afb6b.pdf" }, "presentation": null, "primary_area": { "value": "generative models" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/6a61193071ae2b9548fa31ed5f73a2bbacfea3e4.zip" }, "title": { "value": "Learning Identifiable Concepts for Compositional Image Generation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 4 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "The paper proposes a jailbreak attack against Large Audio Language Models that can enable users to extract harmful information from these models cause them to respond to other users in a harmful manner." }, "flag_for_ethics_review": { "value": [ "Yes, Privacy, security and safety", "Yes, Potentially harmful insights, methodologies and applications" ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Why is noise-to-signal ratio used instead of the more common signal-to-noise ratio? Is it computed in a similar manner as SNR? The normalization and subtraction yields a quantity that is proportional to SNR so perhaps its simpler to just use SNR.\n1. How exactly is $S_{\\text{Mel-Sim}}$ computed? The mel spectrogram is a matrix so how exactly is the cosine similarity computed? \n 1. Why is cosine similarity used instead of L2 distance that is commonly used to compare mel spectrograms? I am not sure if the cosine similarity has a reasonable interpretation for mel spectrograms." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The paper is well written and generally clear\n1. The proposed approach is novel and fills an important gap in the current literature.\n1. The proposed attack is successful on diverse models which indicates its generalizability\n1. Using an audio classification loss to make the adversarial suffix resemble natural sounds is an interesting and novel approach" }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents a gradient-based jailbreak attack against Large Audio Language Models (LALM). The proposed method optimizes an adversarial audio suffix that bypasses the safety alignment of the LALM and causes it to produce harmful outputs. To account for the discretization performed to convert continuous audio representations into discrete tokens, a \"dual-phase\" optimization method is proposed whereby, first, the discrete token sequence is optimized to produce the desired harmful output and then the audio suffix is optimized to yield the discrete audio token sequence. Additionally, an adaptive search procedure is proposed to determine the best target for the adversarial loss optimization, and a loss component is introduced to make the adversarial suffix resemble a given environmental sound. Results show that compared to baselines the proposed approach greatly improves attack success rates." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The -Trans baselines seem to weak because these attacks tend to introduce symbols, like special characters, punctuations and emojis, that are not vocalizable so it is expected that generating speech from them will produce weak results. I recommend presenting results for the text-only attack along with the -Trans attack. This way the actual advantage of exploiting the auditory modality will become apparent.\n 1. A better baseline could be to adversarially attack an ASR model that uses the same audio encoder as the LALM such that the target transcription is the text-only attack string.\n\n1. More details about the human evaluation score ($S_{\\text{Human}}$) are needed, including the number of raters, inter-rater agreement, and did all raters rate all the test audios.\n1. The normalization used for the stealth scores seems to be weight the components unfairly. The NSR and cosine are normalized by their theoretic maximum, while the human score is unnormalized so if the actual NSR and cosine scores occupy a smaller range then their contribution to the score will be penalized. A better normalization scheme might be to normalize the mean to 0.5 and standard deviation to 0.125.\n1. The presentation can be improved:\n 1. Phase II is to the left of Phase I in Figure 1. I suggest reorganizing it to make it appear to the right.\n 1. The phrase \"gradient shattering\" or \"shattered gradients\" is confusing here because in prior work it refers to the specific phenomenon that as neural networks become deeper their gradients resemble white noise [1]. The particular phenomenon of relevance in this study is generally referred to as \"gradient obfuscation\" or \"obfuscated gradients\".\n 1. The phrase \"retention loss\" is confusing because it is not clear what is being retained. The target discrete token sequence can not be \"retained\" because the encoder currently does not output it and it is being optimized to do so. Perhaps, \"alignment loss\" or \"sequence loss\" might be better.\n 1. It is not clear from equation 2 that only a suffix is being optimized. It appears that the entire audio is being optimized.\n\n\n[1] Balduzzi, David, et al. \"The shattered gradients problem: If resnets are the answer, then what is the question?.\" International conference on machine learning. PMLR, 2017." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Did you try other audio-transcribed jailbreak classes, including more naturalistic text like in Zheng et al's persuasion paper? [1]\n2. What made you think GCG and BEAST were strong baselines when translated into audio? \n3. Did you attempt your jailbreaks on any versions of Gemini or 4o? To my understanding some of the more capable models are only trained to recognise speech data - which would presumably make your noise perturbations less effective?\n4. Who were the humans judging your stealthiness? was there a more detailed rubric you can share?\n\n\n\n\n[1] https://arxiv.org/abs/2401.06373" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "This paper tackles a important and as yet unaddressed issue in jailbreak literature, and does so with sensitivity to realism. I am particularly impressed with the authors' operationalisation of stealthiness as urban noise (pending audio samples that I can listen to when the paper is out). The authors' use of human judgment to verify and counterweight their classifier (itself a potentially valuable contribution to audio-jailbreak defense) strengthens my confidence in these results even if I can't aurally verify them myself. \n\nThe results of their optimisation are strong. Their ASR results are comparable to or exceed the other audio-jailbreak papers I know of that were submitted to ICLR.\n\nThe methods developed to optimise jailbreaks against audio-models, given the challenges the authors list, are valuable and novel contributions themselves. In particular the method for adversarial optimisation target search seems to me to strengthen the jailbreak method over the baselines they test against. For example, GCG is reasonably well-known for optimising for outputs such as \"Sure,\" even if immediately followed with \"I can't help with that.\" The adaptivity and greater detail of the jailbreak targets listed in the appendix seem to me to increase the likelihood that jailbreaks listed as successful in this paper do in fact contain jailbroken information. I'm also given increasing confidence in the evaluations of this paper by the authors' use of both a word-based classifier that detects refusal strings, and an LLM graded response." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduce a novel jailbreak framework for optimising audio jailbreaks against audio-language models (ALMs). They overcome challenges in the design of ALMs: namely 1.) they find a dual-phase training process so they can optimise attacks even through discretisation operations in the tokeniser, and 2.) they develop an adaptive search method to find more flexible adversarial targets.\n\nFinally, the authors introduce a realistic constraint on their work: that the audio jailbreaks are stealthy. They operationalise this as having human and ALM-based classifiers independently score the audio input for signs that it was adversarially tampered-with. The authors claim (it's hard without hearing audio samples myself) that their jailbreaks are hence indistinguishable from normal urban noise (e.g. a car horn)." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While I'm overall very positive on this paper, I'm a little underwhelmed by the baselines. I would expect that the adversarial perturbations of GCG and BEAST to be quite brittle to being converted to spoken text and then fed into an ALM. These are worthwhile baselines to run, but more semantically-natural baselines like AutoDAN would have pushed the paper even further. The authors acknowledge the difficulty and novelty of introducing audio-based adaptive attacks, like transfers of PAIR or TAP: I would have been very excited to see the authors tackle adaptive jailbreaks in the audio domain, but understand why for reasons of cost and difficulty that this might not be feasible - though I am aware of an unpublished audio implementation of PAIR.\n\nI think Fig 1 is quite challenging to parse. I would rather it be simplified quite a lot more before final release. In particular, I think there is too much text annotating Phase II, even if helpful for diving deeper into the method. I would prefer at least a much more abstracted version of the figure, without reference to variables from Equation 1, and with the annotation retooled to explain how the different branches refer to each other. At the moment I think it's too hard to understand without continuous reference to Equation 1, and the figure struggles to explain itself on its own." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Have you thought about measuring how your attacks transfer between models? I’d love to see transferability in your work since the threat model I think is most concerning is people finding white-box attacks on open-source models that transfer to more powerful closed-source models. See examples here: https://arxiv.org/abs/2403.09766 , https://arxiv.org/abs/2407.15211\n\nSmall discussion point on using LALMs. Most of the field uses VLMs for vision language models, so do you think using ALMs would be a better acronym to popularise in the field?\n\nI have weaved most of my questions into the weaknesses section. I think this paper has the potential for a much higher rating (especially given the timeliness of getting attacks working on LALMs, which is a neglected area of the adversarial attack literature), but not in its current form. I am happy to increase my score if the weaknesses I highlighted are addressed." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "- **Relevant and timely approach**: LALMs are becoming more prevalent with the recent release of audio capabilities in Gemini and GPT-4o. However, to the best of my knowledge, AdvWave is the first work that has successfully got white-box jailbreak attacks to work.\n- **Innovative approach**: AdvWave uses a dual-phase optimisation strategy to address the issue of not being able to backpropagate through the full network when it contains discretisation. They also improve optimisation efficiency by adaptively finding a target string that matches the common structure the model uses for benign requests. These challenges are clearly explained, and the authors provide solutions.\n- **Potential for future research**: AdvWave opens several avenues for future work, including further exploration of defensive mechanisms and applying the framework to LALMs that do not have a discretization bottleneck." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces AdvWave, a framework for conducting white-box adversarial attacks against large audio-language models (LALMs) to elicit harmful information. The authors identify the unique challenges posed by LALMs, such as gradient shattering due to discretisation operations in audio encoders and maintaining stealthiness constraints. To address these issues, AdvWave implements a dual-phase optimisation strategy. The first phase optimises a discrete latent representation to circumvent the gradient shattering issue, while the second phase adjusts the audio waveform itself to align closely with this representation while preserving perceptual naturalness. AdvWave significantly outperforms transferring static jailbreak attacks optimised on text-only LLMs that are subsequently vocalised with text-to-speech (TTS). The authors argue that their approach highlights the need for more robust defenses and safety measures in LALMs." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- **AdvWave is not the first approach to jailbreaking LALMs:** The authors of the paper claim that AdvWave is a novel framework for jailbreaking LALMs, but I think this claim needs to be softened throughout the paper. Gemini and GPT4o both have audio-specific jailbreaks, where they vocalise text jailbreaks with TTS in their recent model cards. Therefore, claiming that AdvWave is a novel white-box attack methodology for LALMs is better.\n- **Stealthiness constraints are not well-motivated or measured:** It isn’t clear to me why this stealthiness constraint is needed. Any jailbreak, no matter the stealthiness of the input, is bad. Also, stealthiness constraints are not new; they were used in white-box attacks of speech-to-text (STT) models, and the intro doesn’t explain why LALMs make it more difficult. Papers such as “Imperceptible, robust, and targeted adversarial examples for automatic speech recognition” should be cited. Did you ablate changing the whole audio file rather than just the suffix? What motivated the suffix and the environmental noises? You can have imperceptible changes to the audio without that, I believe. Also, what environmental classifier do you use? This needs to be cited for replication purposes.\n - I’m very confused by the stealth metric and why it is a useful comparison to the baselines. The baselines do not have adversarial suffixes added on; they are just TTS reading a harmful prompt. So why isn’t their S_stealth equal to 1? It should be maximally stealthy, just like the vanilla baseline. Also, the baselines do not have a car horn at the end of the utterance, which could be considered more stealthy than your method. You mention you need high stealthiness so it is less detectable by real-world guardrail systems, but I don’t think the results presented demonstrate this. Also, AdvWave is not superior to vanilla in terms of stealth. The three terms in S_stealth are confusing and not well-motivated.\n- **Lack of relevant adaptive black-box baselines:** The paper only compares attacks that are optimised on text-only LLMs that are then transferred into audio with TTS. Using TTS to vocalise GCG attacks might not make sense - there could be lots of tokens that can’t be spoken properly so I would expect the attack to be very weak. You say there are no adaptive attacks due to gradient shattering, but there are plenty of good adaptive black box attacks. I expect PAIR and TAP to work well in the audio domain. AdvWave should be evaluated against much stronger baselines than currently used. How strong are the transfer GCG/BEAST attacks to the base text-only LLM? E.g. what is the GCG ASR on Llama3? That would inform if the baselines transfer to the audio domain effectively or if they are broken by the architectural / fine-tuning differences.\n- **Lack of clarity on LALM architecture differences, what architecture AdvWave is aimed at, and motivation for why dual optimisation to solve gradient shattering is needed:** Not all LALMs have discretisation of audio before input to an LLM (like SpeechGPT). Many insert a continuous vector directly into the embedding space of the model (e.g. DiVA, Llama3.1, Salmonn, Llasm, AudioQwen). Therefore, these won’t have the gradient shattering problem, and the framework in Figure 1 isn’t relevant. There needs to be better motivation and explanation of why AdvWave targets LALMs that have the discrete bottleneck. Ideally, the paper will explain all the different architectures and introduce a framework that works for all the variants. Also, many LALMs do not have a decoder that maps back to audio space. Lots just do audio-to-text. Only a few models are fully speech-to-speech (some are trained end-to-end, and others just put a TTS module on end). It is important to talk about these. Furthermore, why can’t you use a straight-through estimator or Gumbel softmax to differentiate through the discretisation instead of the dual optimisation approach? I need more motivation to believe this is necessary.\n - Also, is gradient shattering a well-known term? A quick search gets something different: https://arxiv.org/abs/1702.08591. Perhaps the problem could just be called “Non differentiable audio tokenisation” or similar? I don’t think the dual optimization method is novel, it would be good to find the original paper that implements something like this. PErhaps it would be in the VQVAE literature?\n- **Lack of threat model:** I’d like to see your threat model go into depth more about why you focus on white-box attacks and why you need stealthiness constraints. E.g., you can just apply existing white-box attacks to text LLMs already and get bad outputs; why do we care about LALM defense when text isn’t solved? Isn’t an attack that elicits a harmful jailbreak that isn’t “stealthy” also a success from the red team’s perspective? Why does it need to be understandable? These can be addressed in your threat model. Also, you mention in related work that LALMs shouldn’t be deployed widely if they are not robust, but releasing them as closed source is fine since you can’t attack with AdvWave.\n- **Presentation of equations, figures, and results needs to be polished:**\n - Figure 1: Phase 1 would be nicer on the left. A brief intuition on what each loss is trying to achieve in the caption would be helpful\n - Section 3.2, in general, is very hard to follow along. L_retent is talked about lot before being explained. Include an intuitive explanation earlier. You introduce the notation for the size mappings of each component, but this makes it more confusing, in my opinion. I would put this in the appendix.\n - Section 3.5 - There is lots of repetition of equations here (e.g. equ 7 is the same as 5 and 6 similar to 1), it would be great if it could be folded into the other sections for conciseness\n - I’m not sure what the perk of having ASR-W is in addition to ASR-L. Often, LLMs are still jailbroken if they say, “I’m sorry,” so I’d expect ASR-W to have many false negatives. It would be good to manually check the false positive rate of ASR-L.\n - Figures 2 & 3 need axes labels and should use a color-blind friendly palette (without gradients). Figure 4 has text that is too small.\n- **Related work is majorly lacking citations and doesn’t contrast with AdvWave:**\n - Add related work to white-box attacks on VLMs - your work is very comparable to how people jailbreak VLMs, e.g., https://yunqing-me.github.io/AttackVLM/ , https://arxiv.org/pdf/2306.13213, https://arxiv.org/pdf/2402.02309. Also, vocalising the request is similar to putting typographic text into images (like FigStep, Images are Achilles Heel of Alignment, Jailbreak in pieces)\n - Add related work to white-box attacks on STT models - this is also very relevant, especially the imperceivable constraints. e.g. “Audio adversarial examples: Targeted attacks on speech-to-text”, “There is more than one kind of robustness: Fooling whisper with adversarial examples”.\n - There are many more papers than I provide here, and I’d recommend doing a proper literature review.\n - LALM section - I would cut the section around concerns of misuse. This should be discussed in the intro. You should cite frontier models like Gemini and GPT-4o advanced voice mode.\n - Jailbreak attacks on LLMs section - you should cite https://arxiv.org/abs/2404.02151\n- **Adaptive target search seems overly complicated:** why did optimising just for “sure” as the first token not work? This works in VLM literature. When comparing to optimizing for “sure”, did you use a prompt like in https://arxiv.org/abs/2404.02151? If not, optimizing for “sure” alone may be much weaker. I’d expect if you did this, the ASR would increase. Essentially, using an “adaptively search optimisation target,” you find a good starting point, but prompting the model to start the response with “Sure, here is…” might mean you don’t need this component. Also, why can’t you find a target string from another jailbroken LLM even if it has a very different structure to the output of the LALM? Shouldn’t gradient-based approaches still be able to change the model to output this?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "In the supplementary materials provided, I am puzzled about adding adversarial noise: 1. The authors mention that the adversarial noise is naturalized using urban environmental sounds as a masking method. However, I can still hear the traditional adversarial disturbances beyond the environmental sounds, suggesting the presence of two types of perturbations, which the paper does not mention. 2. The attack audio samples provided have adversarial disturbances implanted at the end silence segments of the audio, occupying about half the duration of the audio itself. It's unlikely for such a high proportion of silence in most audio datasets, revealing a serious issue: can adversarial attacks unrestrictively zero-pad benign audio ensure attack success? This seems to relate to the authors' initial claim that audio attacks on LALMs would limit the optimization search space for adversarial disturbances. I imagine the authors extended the audio to ensure sufficient search space, yet this seems impractical in real situations. 3. I am curious why the adversarial disturbances were added to the silence segments. Semantically rich portions of the audio seem more susceptible to attacks, and placing disturbances in silent parts would make the noise more detectable by human ears." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The work is groundbreaking as it introduces the first jailbreaking adversarial attack on LALMs. The authors have conducted extensive experimental comparisons, particularly by adapting jailbreaking methods from other domains to the audio sector to ensure the superiority of the proposed attack method. The contribution of this paper is indisputable. However, I still have some questions regarding the audio stealthiness mentioned by the authors." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper presents an innovative adversarial attack method targeting LALMs, marking the first successful attack on LALMs with optimized audio stealth. The efforts are commendable." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I believe the design motivation behind the authors' idea might be flawed. The audio provided clearly contains malicious content, so why consider the stealthiness of the adversarial disturbance? A normal listener would already notice something amiss with the content. Adding adversarial noise to silence segments inevitably leads to listeners hearing malicious content followed by eerie noises, which is utterly unconvincing from a realistic perspective. The authors should more reasonably consider the reasons for the stealthiness of adversarial disturbances and integrate them with the application scenarios of LALMs for a rational design." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We develop the first jailbreak framework against large audio-language models." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024advwave,\ntitle={AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0BujOfTqab},\nnote={under review}\n}" }, "abstract": { "value": "Recent advancements in large audio-language models (LALMs) have enabled speech-based user interactions, significantly enhancing user experience and accelerating the deployment of LALMs in real-world applications. However, ensuring the safety of LALMs is crucial to prevent risky outputs that may raise societal concerns or violate AI regulations. Despite the importance of this issue, research on jailbreaking LALMs remains limited due to their recent emergence and the additional technical challenges they present compared to attacks on DNN-based audio models. Specifically, the audio encoders in LALMs, which involve discretization operations, often lead to gradient shattering, hindering the effectiveness of attacks relying on gradient-based optimizations. The behavioral variability of LALMs further complicates the identification of effective (adversarial) optimization targets. Moreover, enforcing stealthiness constraints on adversarial audio waveforms introduces a reduced, non-convex feasible solution space, further intensifying the challenges of the optimization process. To overcome these challenges, we develop AdvWave, the first jailbreak framework against LALMs. We propose a dual-phase optimization method that addresses gradient shattering, enabling effective end-to-end gradient-based optimization. Additionally, we develop an adaptive adversarial target search algorithm that dynamically adjusts the adversarial optimization target based on the response patterns of LALMs for specific queries. To ensure that adversarial audio remains perceptually natural to human listeners, we design a classifier-guided optimization approach that generates adversarial noise resembling common urban sounds. Extensive evaluations on multiple advanced LALMs demonstrate that AdvWave outperforms baseline methods, achieving a 40\\% higher average jailbreak attack success rate. Both audio stealthiness metrics and human evaluations confirm that adversarial audio generated by AdvWave is indistinguishable from natural sounds. We believe AdvWave will inspire future research aiming to enhance the safety alignment of LALMs, supporting their responsible deployment in real-world scenarios." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "jailbreak", "adversarial attack", "audio-language model" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/99d9042de14ea3d96059ae2e2456e96d6ecfe53d.pdf" }, "presentation": null, "primary_area": { "value": "alignment, fairness, safety, privacy, and societal considerations" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/69c8e46c0a1fe8e52fa279ed68395137f45d6a44.zip" }, "title": { "value": "AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "Q: Did you all empirically assess how much time this k x d^3 adds time-wise compared with just a standard AE? Its an offline computation, but getting a sense of what that tradeoff comes out to time wise for your datasets would be interesting ( ie, is it that big of a hit in the end since the datasets are all below 45k instances each ). How does KernalPCA perform? \n\nQ: line 138 should probably cite the 2017 Transformers paper and not the 2023 arxiv one\n\nQ: Is there a reason in particular for using tanh activations in your extension of CAE? I get it allows for the decomposition shown, but are there other activations or ablations which could have been performed ?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "The authors show how the CAE framework ( an AE with an additional objective component that is the Frobenius norm of the model with respect to the input ) can be extended to a multi-layer setting in a way that is advantageous when compared with prior extensions to CAEs which worked via stacking. \n\nThey do an extensive empirical analysis to discuss reconstruction and downstream accuracy benefits of the method while discussing costs of the method (scaling as the method is cubic with respect to layer size ) and its downstream limitations compared to KernalPCA.\n\nThe setting ( encoding tabular data with multi-modal data types ) is important and they show how its not handled readily or generally by Transformer type architectures." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "In this work the authors extend the Contractive AutoEncoder (CAE) framework for the calculation of the Jacobian \nof the entire encoder in the contractive loss from single-layer to multi-layer settings ( DeepCAE ).\n\n\nEmpirically over tabular benchmarks, the authors show DeepCAE can be leveraged in a general purpose embedding \nframework where embeddings are feed to XGBoost to obtain gains in reconstruction performance and comparable/slightly better performance \ndownstream prediction (classification/regression) performance as compared with various AutoEncoders and Transformer baselines \n( though not when compared with KernalPCA from a downstream performance perspective ). \nThey additionally show the noteable reconstruction performance of DeepCAEs compared with Stacked CAEs ." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1) Time comparisons and particularly error bars needed everywhere ( KernelPCA/StandardAE/DeepCAE ). For your benchmarks are you all running multiple seeds per each? \n\nThe main question ( which the authors have pointed out openly in the paper and for future work ) is if the cost of DeepCAEs is worth the effort? They’ve shown reconstruction is slightly better, but for tasks its pretty comparable to AE and KernalPCA does better (still an interesting finding ). How important is reconstruction loss really for this setup? What is the time complexity of KernalPCA and outside of reconstruction loss being subpar are there other reasons to not use it?\n\n2) Are there stronger baselines to compare against both encoding wise and classifier wise (ie, XGBoost vs something else) against if what we care about is tabular performance using embeddings? The former is the more important of the two and there is a NeurIPS workshop on fusing modalities for tabular data thats in its 3rd edition (https://table-representation-learning.github.io ) \n\n3) The general purpose embedding pipeline seems like the standard solution to re-using embeddings for downtsream tasks from vector databases and not particular to DeepCAE? Is this the case? If not, it could strengthen the paper to clarify how so if not.\n\n4) It would be interesting to either show experiments on or discuss how DeepCAE does on just image or text data as well to compare its reconstruction and downstream task performance there. Is there anything in particular that makes this approach specific to tabular data with multi-modal data? If the method gives performance boosts when encoding image/timeseries/text, it would greatly strengthen the results of the paper and would make incorporating the method.\n\n5) While the background on CAE was very much needed, the sections on VAE and Transformers were probably lesser so ( or could have been pushed into the appendix ) especially since you show effectively they are not nearly as effective. This space could/should be used for looking more at KernalPCA vs Standard AE vs DeepCAE costs/tradeoffs and potentially other modalities ( point 4)\n\n6) Did you all do experiments against the original CAE? The paragraph starting at 280 made it seem like you would ( and in the final section you do with StackedCAE), but then in the experiments Convolutional AE is used instead which I wasn't expecting. I'm assuming this is shown in past work when Stacked CAEs are introduced but having a sense of that as well would be good since its much cheaper computationally than both Stacked and Deep CAEs." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 1 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "In the existing regularization strategies, have other methods been considered, such as dropout or data augmentation? These techniques have been proven effective in preventing overfitting." }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "This paper introduces DEEPCAE, a multi-layer contractive autoencoder designed for general-purpose entity embedding. By extending the contractive autoencoder framework to multiple layers while preserving regularization, DEEPCAE overcomes limitations seen in stacked CAEs, opening new possibilities for autoencoders with high-dimensional data. The study is thorough, with DEEPCAE evaluated across 13 diverse datasets, showing consistently strong results in both reconstruction and downstream tasks that highlight its effectiveness. The paper is well-organized, with clear derivations and detailed appendices on model architecture and hyperparameters to ensure reproducibility. Overall, DEEPCAE offers an efficient, versatile solution for embedding across domains, reducing feature engineering time and adding practical value for cross-application embeddings in industrial settings." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces DEEPCAE, a versatile entity embedding framework based on autoencoders. By extending contractive autoencoders (CAE) to a multi-layer structure and preserving the original regularization design, DEEPCAE enhances both reconstruction accuracy and downstream prediction performance for complex entity embeddings. In tests across 13 datasets, DEEPCAE consistently outperformed other autoencoder variants in both reconstruction error and predictive tasks, achieving a 34% reduction in reconstruction error compared to a stacked CAE. This framework offers an efficient, scalable solution for general-purpose entity embeddings across diverse domains, ultimately reducing time spent on feature engineering and boosting model accuracy." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "DEEPCAE demonstrates impressive results with contractive regularization, but it may not have explored other well-established regularization techniques, like dropout or data augmentation, which are effective in preventing overfitting. It would be beneficial for the authors to consider incorporating these strategies into the DEEPCAE framework. Doing so could enhance the model's robustness, and evaluating their impact on performance in future experiments would provide valuable insights into improving its effectiveness." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "Please see the weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The extension of CAE for multi-layer setting is simple yet effective. \n\nThe ‌authors conduct‌ experiments across 13 datasets and ‌cover‌ various types of entities.\n\nThe results demonstrate state-of-the-art performance on both reconstruction and ‌downstream‌ prediction." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes DeepCAE for learning general-purpose entity embeddings. Although DeepCAE extends from the contractive autoencoder, the authors provide a more effective design in calculating the multi-layered regularization term. Extensive experiments across 13 datasets demonstrate state-of-the-art performance of DeepCAE on reconstruction and ‌downstream‌ prediction tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The main paper should be self-contained. The authors may overly refer to the ‌original‌ CAE paper.\n\nThe motivation of DeepCAE and CAE is not clearly introduced. It is ‌confusing‌ for me why ‌they are designed for tabular data, and how ‌they are connected‌.\n\nIn the experimental results, the strengths of DeepCAE are not significant compared with the standard AE." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 1 }, "primary_area": null, "questions": { "value": "1. The argument \"*Consequently, these representation methods not only are limited, but also fall short or are inapplicable ...*\" in line 47-48, do you have evidence to support this claim, such as references or experimental results? Also, what do \"*these representation methods*\" refer to?\n2. In the following paragraph starting at line 50, how are the two contributions related? The authors might consider clarifying that the framework is based on their proposed method.\n3. In section 2, what is the purpose of elaborating on PCA, variational autoencoders, and transformers? If they only serve as baselines, a brief introduction in the experimental setup might suffice. Instead, the focus should be on elaborating CAE in this section on its principle and structure. For examples, the steps to encode the input and decode it, the loss function.\n4. Eq. (1) lacks detailed description, such as the meaning of its symbols.\n5. In line 92-93, \"*Thanks to their ability to produce stable and robust embeddings, CAE were proven to be superior to Denoising Autoencoders (DAE)*\", the real reason behind *produce stable and robust embeddings* and *superior performance* is missing. The authors should make this argument more well-founded.\n6. In the first paragraph of section 4.1, the sentence \"*we analyze related work and find that, to the best of our knowledge, all use stacking Wu et al. (2019); Aamir et al. (2021); Wang et al. (2020), including Rifai et al. (2011b), who originally proposed the CAE.*\" is poorly written. Additionally, what kind of \"stacking\" was used? The authors should elaborate on how previous works implemented this.\n7. In line 223, how is the conclusion \"$O(d_x \\times d_h^2)$ to $O(d_x \\times d_h)$\" derived? I cannot deduce this from Eq. (3) alone. And \"*and $d_h$ is the hidden embedding space.*\", do you mean $d_h$ is the dimension of the hidden embedding space?\n8. The explanation of how Eq.(4) is obtained from Eq. (3) is unclear. The entire inference process from Eq. (3) to Eq. (8) lacks coherence.\n9. In line 274, \"*such as text and time-series*\" what kind of information is the \"text\" and \"time-series\" exactly? Could you provide examples?\n10. In Figure 1, how exactly do you concatenate the text encoding, TS encoding, and tabular data together?\n11. In section 5, what are the experimental settings, such as the number of encoder layers and feature dimensions? Additionally, the dataset statistics should be included in the main paper rather than the appendix, as they are important.\n12. How is the Mean Squared Error (MSE) computed? Is this metric conventionally used in previous work?\n13. In the first paragraph of section 5.1.2, \"*We trained XGBOOST (Chen & Guestrin, 2016) predictors ...*\", what is the XGBOOST model, and what exactly is the downstream task?\n14. In section 7, line 522-524, \"*Furthermore, we argue that the augmentative capabilities of more complex architectures like Transformers and CNNs are not necessarily useful in the production of a compact representation of an entity.*\", do you have evidence to support this argument?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The authors conducted sufficient experiments to validate the effectiveness of their proposed method. They also provide code and data to reproduce the results." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes a method called DeepCAE to calculate the regularization term for multi-layer contractive autoencoders and utilizes DeepCAE to power a general-purpose entity embedding framework. The experimental results show that DeepCAE outperformed the baselines." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "I think the prominent issue is the writing. 1) It **contains lots of tangential content** which is unnecessary to elaborate on in this paper. For example, in Section 2, I don't see the point of discussing PCA, variational autoencoders, and transformers, as they are not directly part of or foundational to your methodology. 2) The writing **lacks substantial details and is not self-contained**. This work is largely based on the previous CAE work by Rifai et al. (2021b), but the introduction to that work is incomplete, making it difficult to understand the proposed method and its details. Instead, the authors frequently refer readers to the original work. This problem also exists in the experiment, which did not clearly present the experiment settings. 3) **Lacking of logical coherence**. Some important arguments lack evidential support and are not clearly described.\n\nPlease see questions below for exact points." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We introduce DeepCAE, an enhanced multi-layer contractive autoencoder, and benchmark autoencoder architectures in a general-purpose tabular data embedding framework for reconstruction and downstream performance, with a 34% reconstruction improvement." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024autoencoderbased,\ntitle={Autoencoder-Based General-Purpose Representation Learning for Entity Embedding},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0C5iHPPwsG},\nnote={under review}\n}" }, "abstract": { "value": "Recent advances in representation learning have successfully leveraged the underlying domain-specific structure of data across various fields. However, representing diverse and complex entities stored in tabular format within a latent space remains challenging.\nIn this paper, we introduce DeepCAE, a novel method for calculating the regularization term for multi-layer contractive autoencoders (CAEs). Additionally, we formalize a general-purpose entity embedding framework and use it to empirically show that DeepCAE outperforms all other tested autoencoder variants in both reconstruction performance and downstream prediction performance. Notably, when compared to a stacked CAE across 13 datasets, DeepCAE achieves a 34% improvement in reconstruction error." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "customer", "embeddings", "embedding", "tabular", "general", "purpose", "autoencoder", "representation learning", "general purpose", "reconstruction loss", "entity", "entity embedding", "entity representation", "contractive autoencoder", "dimensionality", "reduction", "latent", "space", "representation", "feature", "regularization", "variational autoencoder" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/82d6ecb8781f42a98cd47903bf699fdc72d681e0.pdf" }, "presentation": null, "primary_area": { "value": "unsupervised, self-supervised, semi-supervised, and supervised representation learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/20d1da13784f315c8ad4e8924a1cfa9107f69b98.zip" }, "title": { "value": "Autoencoder-Based General-Purpose Representation Learning for Entity Embedding" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- The discussion of the different activation functions (appendix) is indeed interesting. And this could be one of the interesting parts of the ablation study. However, it is strange to see that using the Gaussian non-linear function is yielding very bad performance. Perhaps the spectral width needs to be fine-tuned, especially when the distribution of the lidar scene flow is very unique.\n\nPlease also see the above section for detailed comments." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- The paper proposes an interesting idea to represent the scene flow as a velocity field using a neural network, making it very easy to combine the temporal information (time) and the spatial information (position of points).\n- The authors have done extensive analysis of the proposed method, and have shown different ablation studies to validate the effectiveness of the method.\n- The proposed method also shows the potential to deal with small objects and emergent flows in robotics scenarios, which could be interesting when applied to highly dynamic environments.\n- Overall, the writing of the paper is clear, and the visualization is easy to understand." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes to represent scene flow as a velocity field using a neural prior. Instead of prior art that directly represents per-pair scene flow as neural prior, the authors alternatively propose to use neural prior to model the partial differential equation (PDE) of the position of the point versus the time interval. This novel velocity representation is interesting and could offer flexibility in dealing with long-term sequences of flow estimations as the authors described in the paper. The authors have also done extensive analysis of the proposed method on Argoverse 2 (and Waymo) datasets, comparing the performance with recent scene flow works, and validating the good performance of the proposed method." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- When using the time interval between [-1, 1] for the time encoding, will the proposed method not be able to handle time step outside the range? Given that the representation is a continuous neural network, how does it extrapolate to a longer sequence with the current representation?\n- When comparing with a method like NSFP, I wondered if the authors could show the results of pure Euler integration of the method and highlight the benefits of wrapping a PDE with a neural network.\n- The authors mentioned that they only do sequences of length 20, I wondered if the method failed rapidly with the increase of the sequence length. It would be interesting to show an even longer sequence to highlight the arbitrary time query property of the proposed method.\n- I feel like the authors want to talk about too many things in this paper, so they may overlook the most important part of the method. This method is good at dealing with long-term flow trajectory and has the potential to better capture the small, highly dynamic objects in the scene. The authors could reorganize the motivations and experiments to highlight the advantages of the proposed method." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Overall I believe this paper is in a good shape, the authors discuss the properties and limitations of the proposed method thoroughly in the paper. I have a few more questions:\n\n- How does the method handle scenes with deformable objects?\n- What is the impact of temporal sampling rate on performance?\n- How does the point cloud density affect the performance?" }, "rating": { "value": 8 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- I'm not up-to-date to the latest scene flow models, but from the results in the paper it surpass the prior art by a large margin, which is very significant\n- Introducing the concept of modeling scene flow as a PDE is innovative and offers a new direction for research in motion estimation.\n- The method is rigorously developed, with comprehensive experiments and ablation studies that validate the approach.\n- The paper is well-written, with clear explanations and effective use of figures to illustrate key points." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a novel approach to scene flow estimation by reframing it as the task of estimating a continuous space-time partial differential equation (PDE) that describes motion across an entire observation sequence. The proposed method, called EulerFlow, utilizes a neural prior to represent this PDE and optimizes it against multi-observation reconstruction objectives. This approach allows for high-quality, unsupervised scene flow estimation on real-world data, outperforming existing supervised and unsupervised methods, particularly on small and fast-moving objects like birds and tennis balls. The authors demonstrate the effectiveness of EulerFlow on datasets such as Argoverse 2 and Waymo Open Dataset, and show its applicability across different domains without domain-specific tuning. Additionally, they highlight emergent 3D point tracking behavior as a result of solving the estimated PDE over long time horizons." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- As stated in the paper, the speed of the proposed method is a big concern, preventing it from deploying on real world application.\n- Some hyperparameters, such as the depth of the neural prior, seem to require dataset-specific tuning (e.g., increasing depth to 18 for the Argoverse 2 challenge), which may affect the method's out-of-the-box applicability.\n- It would be great if the author could show more failure cases to help readers better understand its limitations." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Among all the three points I illustrated in the weakness, \n\nFor point 1, I hope the authors can provide concise update on their paper title and contributions in particular for the first bullet time (line 99-100).\n\nFor point 2, the current evaluation is sound and maybe sufficient for this paper. I do believe it is nice to more evaluation on non-AV dataset quantitatively that very likely will benefit this method as a baseline for future work in different domain. \n\nFor point 3, it will be good if the author can provide specific example (as a figure)" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The proposed scene flow representation is simple and technical sound. Compared to prior work NSFP, extending it to multi-frame and learns a bi-directional consistency scene flow is a very intuitive step forward. \n2. The performance of this method (both qualitative and quantitative) is impressive. The method can learn very consistent scene flow in trajectory despite not explicitly considering common issues such as occlusion artifacts. As the paper demonstrated, it can tackle well on small objects (with potentially large motions as well)." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes a neural representation to optimize scene flow as a discrete partial differential equation of scene geometry over time. Compared to previous method, Neural Scene Flow Prior (NSFP), a method is most related to this work in the neural representation, the proposed method introduces a multi-frame formulation and learns bi-directional three-step Euler integration of the geometry consistency using decoded per-frame scene flow. Compared to previous work, the proposed representation can achieve better performance in Autonomous driving datasets and the authors demonstrate qualitative performance on depth camera input as well." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The paper title and introduction is very general and does not provide a precise position of this paper's main contribution. \"Scene flow a partial differential equation\" has been historically formulated long time ago in many prior paper, e.g. [1] as one examplar reference, and it has been proposed as a continuous representation in one early seminal work [2]. Many related work studied this optimization problem using images input and solved it using differential optimization approaches before. In this paper seems only consider related work in the point cloud space, and beneficially solved in using a neural representation. I will suggests to more precise position their scope and contributions in paper title, introduction and contributions. \n\n2. The evaluation dataset in this paper is mostly on autonomous driving datasets though as the method demonstrated, it should also work on other data domain when depth is available. Though real world depth and flow ground truth is hard to get, it won't be too hard if evaluated using a more general synthetic dataset that provide different motion patterns, compared to the type of motion and accuracy that autonomous driving dataset can provide. \n\n3. The paper has already discussed the main limitations it section 6.1. Particularly for the last point \"EulerFlow does not understand ray casting geometry\", it was clear how this has been demonstrated in the current results. It will be good if the authors can provide examples and metrics that reflect the challenge in this aspect.\n\n[1] A Variational Method for Scene Flow Estimation from Stereo Sequences, Frederic Huguet and Frederic Devernay, ICCV 2007\n\n[2] Three dimensional scene flow, Vedula et al, CVPR 1999" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Here are some questions and concerns regarding the presentation and the method:\n\n1) In lines 189 and 195 $\\frac{\\partial L^*}{\\partial t}$ is referred to as the partial differential equation, or a PDE. However, $\\frac{\\partial L^*}{\\partial t}$ alone is not a PDE yet, unless it is set equal to something (as in equation 2).\n2) I guess that in equation 2 SFvPDE should also depend on $x$. Could the authors clarify this?\n3) In general, it would be nice to have more formal definitions. E.g. in EulerFlow, an exact formula for solving the PDE, $\\text{Euler}_\\theta(P_t, d, k)$, would improve understanding and reproducibility of the method.\n4) In principle, the PDE can be integrated in both directions by simply reverting the time. The usage of the direction as an extra argument in the model makes the connection between sections 3 and 4 slightly weaker and seems to be a legacy design choice from NSFP. Thus, a question to the authors is whether they have tried training without the direction argument?\n5) Given the high computational complexity of the method, it would be better to see some implementation details on how exactly equation 3 is calculated during training. Are any optimizations already incorporated? E.g. in the current form separate terms in the loss are independent. However, I believe that subsequent Euler steps can use previous estimates instead of recalculating them.\n6) More ablation studies would better highlight the contributions of the paper. E.g. how general and how sensitive is the method to different numerical solvers and sizes of discretization steps? Have the authors tried higher-order PDE solvers or using $\\Delta t$ smaller than the time between observations?\n\nI will adjust my score based on the other reviews and the rebuttal by the authors." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The method is simple and intuitive yet effective. Extensive experiments section shows clearly that the proposed method surpasses prior work." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper proposes SFvPDE, a framework to cast scene flow estimation as a PDE with a neural prior, and EulerFlow, an example demonstrating how SFvPDE can be trained using the Euler method to locally integrate the PDE during training. A space-time-dependent vector-field is trained to match subsequent point clouds at different timestamps via solving the underlying PDE. The method significantly outperformes both supervised and unsupervised baselines and is especially effective on small objects compared to prior work." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The main weakness of EulerFlow, as also noted by the authors (lines 524-528), is the time it takes to converge on a single scene. But given the performance of the method, this should not be considered critical. However, the presentation of the paper can be improved, the paper lacks some implementation details, as from time to time the reader has to guess what is actually happening (see questions section)." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We model scene flow as an estimating a PDE over many observations; our unsupervised method is high quality (SotA on important benchmarks) and works out-of-the-box on many diverse domains." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024scene,\ntitle={Scene Flow as a Partial Differential Equation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0CieWy9ONY},\nnote={under review}\n}" }, "abstract": { "value": "We reframe scene flow as the task of estimating a continuous space-time PDE that describes motion for an entire observation sequence, represented with a neural prior. Our method, _EulerFlow_, optimizes this neural prior estimate against several multi-observation reconstruction objectives, enabling high quality scene flow estimation via pure self-supervision on real-world data. EulerFlow works out-of-the-box without tuning across multiple domains, including large-scale autonomous driving scenes and dynamic tabletop settings. Remarkably, EulerFlow produces high quality flow estimates on small, fast moving objects like birds and tennis balls, and exhibits emergent 3D point tracking behavior by solving its estimated PDE over long-time horizons. On the Argoverse 2 2024 Scene Flow Challenge, EulerFlow outperforms _all_ prior art, surpassing the next-best _unsupervised_ method by more than 2.5x, and even exceeding the next-best _supervised_ method by over 10%." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Scene Flow", "Neural Prior", "Partial Differential Equation", "Reconstruction" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/e02966061f51590650ee53716d5eba9c7c74b46f.pdf" }, "presentation": null, "primary_area": { "value": "applications to computer vision, audio, language, and other modalities" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/627bb7b18ca6132faec34f993b20b47575f5ef5d.pdf" }, "title": { "value": "Scene Flow as a Partial Differential Equation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "None" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Minor:\n1.\tThe type function used for S_t is not specified. I suppose it is a step function, if so could you please confirm and specify?\n\n2.\tWhere does the name \"Artsy\" originate from? Is it an abbreviation for \"ARTificial Synapse\"? If so, could you specify this in the paper?\n\n3.\tCan the Artsy framework work in other continual learning settings beyond class-incremental learning (CIL)?\n\n4.\tHow are both the pre-trained network and the initialized subnetworks analogous to the mature brain network? Could you provide examples and references of networks and subnetworks coexisting and connected in the mature brain, but with different dynamics?\n\n5.\tWhat is the rationale behind naming the connections \"artificial synapses\"?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. The paper presents an interesting and original idea by leveraging biological mechanisms of learning to enhance AI models, specifically focusing on the activation mechanisms of silent synapses through spike-timing-dependent plasticity. Given that the biological brain exhibits minimal effects of catastrophic forgetting compared to AI models, seeking inspiration from neurobiological learning mechanisms is a promising and innovative research direction. While initial explorations exist in the literature, there remains substantial room for further innovative research in this area.\n\n2. Although some additional details are necessary for a complete explanation and reproducibility of the experiments, the authors have made a commendable effort in describing the framework by providing both training and inference algorithms and biological motivation.\n\n3. The results, while needing a few more details for complete clarity, appear promising and suggest that the Artsy framework outperforms conventional methods on class-incremental learning tasks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces the Artsy framework, which enhances continual learning in pre-trained models by mimicking the activation mechanisms of silent synapses via spike-timing-dependent plasticity observed in biological neural networks. The framework maintains memory stability for previously learned knowledge in the pre-trained network while promoting learning plasticity in task-specific sub-networks during training. During inference, it uses artificial silent and functional synapses to connect pre-synaptic neurons in the pre-trained network with post-synaptic neurons in the sub-networks, enabling effective information extraction. Experimental results show that Artsy outperforms conventional methods on class-incremental learning tasks and offers better biological interpretability than other solutions to mitigate catastrophic forgetting." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "•\tThere are other literature works that propose biologically inspired solutions to mitigate catastrophic forgetting (see below), which are not covered in the background nor related work sections. I suggest adding these references to provide a more comprehensive context for the proposed framework in the neuroscientific context.\n\n•\tSome parameters needed for reproducibility of the results (incl. number of parameters, type of connectivity e.g. for silent and functional synapses) are not reported. For instance, the paper does not mention how many artificial synapses are used for each subnetwork or whether there is e.g. all-to-all vs sparse connectivity.\n\n•\tThe study mentions limitations of other algorithms in the background section regarding efficiency and computational time. However, the authors do not discuss these features of the Artsy framework compared to other algorithms. Efficiency, model complexity and computational time are important aspects that the authors should quantitatively analyze (or at least provide estimates for) to explain the performance vs efficiency tradeoff.\n\n•\tA potential limitation of the framework is the potential increase in the number of subnetworks and artificial synapses with the addition of more classes. This could pose scalability issues and raise questions about the biological plausibility of the framework. I suggest that the authors provide more information and comments on this aspect.\n\n•\tThe paper does not provide a link to the code, which is essential for reproducibility and further validation of the results.\n\n•\tStandard deviations for the results in Table 1 and Figure 3 and 4 are not shown. These are important for comparing the variability of the frameworks. In addition, the experimental setup lacks clarity regarding the number of runs averaged (for instance, for Figure 4B).\n\n•\t“Good” and “bad features” (sec 4.5) are not clearly defined\n\n•\tThe paper makes a claim that brain lesions causing synaptic disconnections can lead to dementia by disrupting synaptic consolidation but lacks references to support this. Moreover, the connection between artificial synapses of the Artsy framework and brain lesions needs to be made clearer.\n\n•\tMore targeted explanations of AMPA and NMDA receptors are needed, for example if the relevance of these receptors to short- vs long-term plasticity is related to the dynamics of functional and silent synapses \n\n•\tThe diagram for Figure 2C could show more than one subnetwork to accurately represent the architecture.\n\nIf the above points are clarified, I am happy to revise my score.\n\nSuggested references (non exhaustive):\nhttps://arxiv.org/pdf/2405.09637 , https://www.nature.com/articles/s42256-023-00747-w , https://www.nature.com/articles/s41467-022-29491-2 , https://arxiv.org/pdf/2403.13249 , https://proceedings.mlr.press/v232/madireddy23a.html , https://pubmed.ncbi.nlm.nih.gov/37589728/ , https://proceedings.mlr.press/v162/gurbuz22a/gurbuz22a.pdf" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. How $thr_t$ is selected? Is it learnable?\n2. What is $h$ in Eq. (5)? Is it $h_0$?\n3. How is $S_t$ optimized (LINE 283)?\n4. How often $m_t=0$ or $m_t=1$? \n5. What does `complete the prototypes for former classes' mean?\n6. How was it determined that the sub-network is trained for 20 epochs and the artificial synapse is trained for 2 epochs?\n7. What is a *good feature*? What is a *bad feature*?\n8. It would be interesting to see the performance on class incremental learning when only the pre-trained model is used. Are such experiments available?\n9. Will the code be publicly available?" }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "**Originality:** An innovative biologically inspired approach to avoid catastrophic forgetting while using pre-trained models for incremental class learning is presented. The emulation of silent and functional synapses to artificial networks is appraising and novel in the context of pre-trained models.\n\n**Quality:** A solid background on the biological foundations necessary to understand the composition of the Artsy framework is given.\n\n**Clarity:** The flow of the text is easy to follow. The objective of the paper is clearly stated. All necessary background is given to understand the presented approach. The experimental setup is explained in detail.\n\n**Significance:** The presented work is significant to advance the research in the domain of continuous/lifelong machine learning. The biological inspiration highlights this." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper addresses the problem of catastrophic forgetting of pre-trained models by presenting an architecture for class-incremental learning.\nThe architecture, the Artsy framework, is inspired by the plasticity of neurons in the brain. \nArtsy simulates silent and functional synapses. Specifically, (1) the fixed pre-trained network acts as a consolidated memory, (2) the sub-network learns the features of new incrementally available data, (3) artificial synapses interconnect the pre-trained network and sub-networks. \nHere, artificial synapses represent the silent and functional synapses.\nThe experimental results show that Artsy achieves superior performance on incremental learning tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "**Inconsistencies in formulas** There are some inconsistencies between equations and the presented algorithms. For example, LINE 230, Eq. (1) defines $h_0 = E_0(x)$, while LINE 272, Algorithm 1, uses $F(x)$ and LINE 326, Algorithm 2, uses $E_0(x)$.\nLINE 323 states $E_0(x) + \\sum_{i=1}^{t}E^i(x) * m_i$ which is different from the expression within the parentheses in Eq. (5). Equation (3) and (4) are not explained enough. For example, what is the purpose of $m_t$ in general (apart from determining whether a synapse is silent or functional) and how $c_t$ is learned?\n\n**Weak ablation study**\nThe ablation study uses two different types of features as input to test the performance. For the ablation study per se, for example, it would be meaningful to see the separate contribution of $E_0(x)$ and $\\sum_{i=1}^{t}E^i(x) * m_i$ to the performance on the class incremental learning task.\n\n**Limited related work** While related work on silent and functional synapses and other approaches for class incremental learning is thoroughly presented, the related work on similar biologically inspired architectures is missing. Here, the comparison with other biologically inspired approaches would be beneficial. As an example, [1] can be considered.\n\n[1] German I. Parisi, Ronald Kemker, Jose L. Part, Christopher Kanan, and Stefan Wermter. Continual lifelong learning with neural networks: A review. Neural Netw., 113(C):54–71, May 2019." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Please see comments for \"weaknesses\"." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The authors propose an automatic gating process that can turn on or off individual adaptors depending on present inputs. For each adaptor, a distinct MLP is trained to predict if a present input is an in-distribution example. During inference, adaptors are activated only when the corresponding MLPs predict “match”. Consequently, if all MLPs are perfectly trained and are 100% accurate, only a single adaptor trained for a present input will be activated, and other adapters will be shut down, which means that we can expect a highly accurate prediction." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors proposed a continual learning system inspired by dynamical switching between silent and functional synapses in the brain. The actual mechanisms, however, has no real link to biological synapses that is extensively discussed in their study. Instead, their algorithm is a variation of the earlier algorithm referred to as ‘EASE’ in this study. EASE uses a pretrained encoder (visual transformer) as a backbone and trains adaptors to learn down-stream tasks. As each adaptor learns a new distinct task, catastrophic forgetting can be avoided." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "This proposed gating process is interesting, but the authors’ own comparison to EASE show that its advantage is marginal. As they used two simple tasks (CIFAR100 and TinyImageNet) to evaluate the newly proposed algorithm, it remains unclear whether the proposed gating mechanism is beneficial for more complex tasks.\n\nAs MLPs need to be trained with old and new data, the algorithm proposed in this study requires a type of replay memory, which is not clarified in the paper.\n\nThe authors’ description of the model (e.g., encoder (E_t(x)), prototypes and MLP) also needs improvements. Their study is based on Zhou et al (2024) study that proposes EASE, so the details of their study may overlap with Zhou’s study, but this does not mean that they do not need to explain their algorithm. They should extend and improve the description of their proposed algorithm for better readability." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Although the methods section of this article is described very clearly, many details are not introduced. For example, the pre-trained network E0(⋅).\n2. The author needs to explain the mathematical mechanisms underlying artificial synapses.\n3. The experiments are too weak, relying solely on two commonly used datasets (CIFAR-100 and TinyImageNet).\n4. The author should include additional experiments that provide interpretability to highlight the advantages of the biological mechanisms.\n5. The author should also provide some efficiency metrics to demonstrate the superiority of the model." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The author's approach to constructing a pre-trained model inspired by the activation mechanisms of silent synapses is commendable.\n2. The overall readability of the article is strong and easy to follow.\n3. The use of artificial silent and functional synapses establishes precise connections between networks, enhancing the extraction of relevant information during inference." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors introduced the Artsy framework, designed to enhance the continual learning capabilities of pre-trained models, addressing their vulnerability to catastrophic forgetting when incrementally trained on new tasks. Using their framework, the authors are able to achieve state-of-the-art performances on class-incremental learning tasks. Furthermore, this framework offered a promising avenue for simulating biological synaptic mechanisms." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The overall experiments have some shortcomings, as only two common datasets, CIFAR-100 and TinyImageNet, were used.\n2. Although the authors emphasize biological synaptic mechanisms throughout the paper, corresponding results are not observed in the results section.\n3. The authors mention that pre-trained artificial neural networks lack generalization capabilities, but they do not conduct corresponding experiments to address this issue.\n4. We would have appreciated a more detailed exploration of how the authors intend to enhance the model based on synaptic mechanisms, accompanied by a mathematical description of these processes. Regrettably, the current explanation remains overly simplistic." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024braininspired,\ntitle={Brain-inspired continual pre-trained learner via silent synaptic consolidation},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0CtIt485ew},\nnote={under review}\n}" }, "abstract": { "value": "Pre-trained models have demonstrated impressive generalization capabilities, yet they remain vulnerable to catastrophic forgetting when incrementally trained on new tasks. Existing architecture-based strategies encounter two primary challenges: Firstly, integrating a pre-trained network with a trainable sub-network complicates the delicate balance between learning plasticity and memory stability across evolving tasks during learning. Secondly, the absence of robust interconnections between pre-trained networks and various sub-networks limits the effective retrieval of pertinent information during inference. In this study, we introduce the $\\textit{Artsy framework}$, inspired by the activation mechanisms of silent synapses via spike-timing-dependent plasticity observed in mature biological neural networks, to enhance the continual learning capabilities of pre-trained models. The Artsy framework integrates two key components: 1) During training, the framework mimics mature brain dynamics by maintaining memory stability for previously learned knowledge within the pre-trained network while simultaneously promoting learning plasticity in task-specific sub-networks. 2) During inference, artificial silent and functional synapses are utilized to establish precise connections between the pre-synaptic neurons in the pre-trained network and the post-synaptic neurons in the sub-networks, facilitated through synaptic consolidation, thereby enabling effective extraction of relevant information from test samples. Comprehensive experimental evaluations reveal that our model significantly outperforms conventional methods on class-incremental learning tasks, while also providing enhanced biological interpretability for architecture-based approaches. Moreover, we propose that the Artsy framework offers a promising avenue for simulating biological synaptic mechanisms, potentially advancing our understanding of neural plasticity in both artificial and biological systems." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Continua learning; Silent synapse; Pre-trained model; neuroscience-inspired method" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/b07e76de0d2f605fe68807d9d8cc11a53e2574f8.pdf" }, "presentation": null, "primary_area": { "value": "applications to neuroscience & cognitive science" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Brain-inspired continual pre-trained learner via silent synaptic consolidation" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Q1. The discussions provided in section 3.2 are not really theoretical analysis. The section title is a bit misleading. I would suggest to either break this section down to the motivation of the work, or rename it to some candidates like intuitional justification.\n\nQ2. In line 340, why only 5 datasets from the UEA is selected, out of 30+ multivariate datasets? Also, the five datasets are finally precessed into univariate datasets, in which case why the original 100+ univariate datasets are excluded?\n\nQ3. The accuracy reported in table 2 is pretty low on the first two datasets. And the differences with or without PCA are huge on some cases, e.g., FEDformer and TimesNet on SelfRegulationSCP1. Could the authors provide justification for these numbers? Otherwise, this would damage the versatility of the proposed approach.\n\nQ4. What is the backbone model for the results in Table 5?\n\nQ5. I am not sure if I fully understood line 302-304, “For example, if all positive trends…” Could the authors further explain a bit on this for me? Thanks!" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "S1. This paper attempts to improve the efficiency of time series analysis tasks, which can be very useful for resource-constrained scenarios including edge computing.\n\nS2. The evaluation was conducted on three different tasks, i.e., time series classification, forecasting, and extrinsic regression, demonstrating the versatility of the proposed methods.\n\nS3. The paper is easy-to-parse." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This manuscript revisits Principal Component Analysis (PCA) to explore its utility in reducing the temporal dimension of time series data, as a novel area of focus, because PCA has traditionally been applied mostly on the variable space. The paper posits that PCA, when applied to sliding series windows, not only maintains model performance but also enhances computational efficiency. Extensive experiments across time series classification, forecasting, and extrinsic regression tasks substantiate these claims. The paper suggests that PCA preprocessing results in reduced training and inference times without compromising on model effectiveness across various deep learning-based time series analysis models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "W1. It is still hard to conclude from this paper that PCA before feeding into deep neural networks is a versatile solution that should be suggested to time series analysis tasks under resource constraints. Briefly speaking, neural networks, especially the early layers of neural networks, are considered as feature extraction as well as dimension adjusting. This overlaps a bit with the purpose of PCA. \n\nW2. There are more dimensional reduction techniques for time series or high-dimensional vectors, rather than PCA itself. For example, DWT, FFT, etc. Although have been briefly discussed, it would still be very important to compare PCA with other deep model-agnostic dimension reduction techniques." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Q1. Could you please include a comparative analysis section in their paper, directly comparing PCA's performance, computational efficiency, and memory usage against the methods you mentioned (T-Loss, TS2Vec, and autoencoder-based approaches). This would provide a clearer context for evaluating PCA's contribution relative to recent advances in the field.\n\n- Q2. Coud you please include state-of-the-art classification models like InceptionTime and ResNet in their comparison for the classification tasks, providing a stronger baseline for evaluating PCA's impact.\n\n- Q3. I Believe that it would be valuable to expand the forecasting experiments to include additional widely-used datasets such as traffic and electricity, alongside the ETT datasets. Suggest specific datasets that are commonly used in the field for benchmarking." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 1 }, "strengths": { "value": "- S1. I think it is interesting to use unsupervised learning as a first step to reduce memory and computation time for downstream supervised tasks. This approach could be particularly beneficial when dealing with large amounts of time series data with numerous timestamps.\n\n- S2. The paper is well written and easy to follow, making the concepts and methods presented clear and accessible to the reader.\n\n- S3. The experimental results show that PCA accelerates both the training and inference processes while reducing the GPU memory usage for the considered downstream tasks." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper explores the application of Principal Component Analysis (PCA) for dimensionality reduction of the temporal dimension. The authors argue that PCA's ability to reduce dimensionality enables the extraction of essential features underlying time series, thereby improving the efficiency of downstream tasks. Experimentally, the study applies forecasting, classification, and extrinsic regression tasks to the PCA-learned representations. The results show significant improvements in computational time and memory consumption compared to purely supervised approaches applied directly to the raw time series." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- W1. A significant weakness of the paper is its lack of discussion and comparison with other representation learning methods. \n - Several claims in the paper appear to be inaccurate, such as: \"To the best of our knowledge, there has been no systematic method for compressing time series data in the temporal dimension while preserving critical information\" and \"far less attention has been given to reducing the temporal dimension, despite the potential benefits of alleviating the burdens associated with processing long time series.\" In recent years, various unsupervised time series methods have effectively addressed this issue. For example, T-Loss [1] was one of the first models to fully compress the temporal dimension by leveraging contrastive learning and an encoder-only architecture. Another contrastive method, TS2Vec [2], learns representations that can be used for forecasting and classification in subsequent stages. Additionally, methods based on autoencoders with vector quantization [3,4] have demonstrated the ability to compress the temporal dimension by learning the core features of time series data. \n - The use of PCA representation does not appear to enhance the performance of the supervised model. While the authors argue that PCA representation accelerates training and inference (and reduces memory usage), the omission of other representation learning methods—such as a basic convolutional encoder-decoder—makes it difficult to fully evaluate the contribution of this paper.\n\n\n- W2. From an experimental perspective, several aspects seem questionable.\n - For the classification tasks, the authors selected a few datasets from the UEA and applied PCA pairs with models that are primarily known as forecasting baselines (except for TimesNet). The reported results, whether with or without PCA, do not represent state-of-the-art performance. It would have been beneficial to include models like InceptionTime or a simple ResNet for comparison. \n - In the forecasting tasks, the authors focused solely on the ETT datasets, which are recognized for their difficulty in forecasting. It would be more insightful to conduct similar experiments on datasets such as traffic or electricity, which may provide additional context and validation for the proposed methods.\n\n\n[1] Unsupervised scalable representation learning for multivariate time series, Neurips 2019\n\n[2] Ts2vec: Towards universal representation of time series, AAAI 2022\n\n[3] Vector Quantized Time Series Generation with a Bidirectional Prior Model, AISTATS 2023\n\n[4] Interpretable time series neural representation for classification purposes, IEEE DSAA 2023" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "1. The paper does not seem to demonstrate whether the number of time steps is reduced or to what extent. It only generally mentions at the beginning that applying PCA will compress the time series, but it is unclear whether the compression target is the time steps or the data component features within the sliding window. In the Introduction, the authors state that dimensionality reduction techniques for time series data mainly focus on the variable dimension, and they intend to apply PCA for dimensionality reduction along the time dimension. However, from the overall description, the authors appear to only apply PCA to the time-step data within the sliding window to extract local features. This method extracts feature information from each window, but the number of time steps within the window seems to remain unchanged. \n\n2. It is currently unclear whether the authors also applied PCA to the test set in the classification task. If the authors used PCA to preprocess the test set, this would be unreasonable because the test data should be assumed to be unknown beforehand. If the authors did not apply PCA to the test set, maintaining the original data format and attributes while keeping the network unchanged, then theoretically, there should not be a significant acceleration in inference time. \n\n3. There is an issue with unreasonable descriptions in the related work section. The authors discuss the limitations of Xu et al.'s work titled \"Transformer multivariate forecasting: Less is more?\" In their second point, they state: \"Secondly, it is designed for scenarios where a multivariate series forecasts a univariate series, focusing on reducing the variable dimension of covariate series without preprocessing the target variable series, even if the covariate series may have minimal association with the target series.\" \n\n4. In the theoretical analysis section, as shown in Figure 3, the authors only demonstrate the effectiveness of PCA in reducing dimensionality along the feature dimension. However, they do not address dimensionality reduction along the time dimension (i.e., the compression of the number of time steps). \n\n5. It appears that the experimental results presented in Figure 3 are based on a single experiment. The authors did not verify the generalizability of their results by experimenting with different numbers of principal components, k. Without varying k, there's a risk that the chosen value might be a \"lucky number\" that coincidentally yields favorable results. \n\n6. The experimental results may be influenced by the characteristics of the specific dataset used. The smoothing effect observed in Figure 3 might only be applicable to the current dataset and may not represent the performance on other time series data. Including experiments on a variety of datasets could improve the credibility of the conclusions. \n\n7.The authors propose that \"Specific trends and periodic patterns in historical series may not be crucial for the learning of TSA models.\" In the field of Time Series Analysis (TSA), traditional viewpoints and a large body of research emphasize the importance of trends and periodic patterns. These elements are critical for understanding the inherent structure of the data and for predicting future values. \n\n8. The authors' statement, \"Therefore, although PCA may alter the trend or periodicity, it introduces new coherent patterns—such as the main directions of variation, denoised low-dimensional representations, and latent features—that benefit TSA model learning without negatively impacting predictive performance,\" is overly absolute. Claiming that there are no negative impacts is too definitive. In practical applications, any data transformation can potentially have both positive and negative effects on model performance; the specific outcome depends on the characteristics of the data and the type of model employed.\n\n9. It seems that the experiments presented in Tables 2, 3, and 4 are based on single runs without any statistical significance testing. There is no indication of whether the results are consistent across multiple trials or if they could be due to random chance. Furthermore, in each experiment, the number of principal components (k) selected for PCA is based on a single value, and this value differs across different datasets.\n\n10. In Table 2, concerning the Time Series Classification (TSC) experiments, the authors conclude based on the results: \"These results reveal PCA’s efficacy in extracting series information for TSC tasks without performance loss, enabling faster training/inference.\" However, the results presented in Table 2 indicate that applying PCA on certain datasets and networks can lead to significant performance degradation. For example, on the SelfRegulationSCP1 dataset, the accuracy of the TimesNet network decreased by 23.2% after applying PCA. This substantial drop contradicts the authors' absolute assertion of \"without performance loss.\" Out of the 20 metrics reported, only 10 show performance improvement when PCA is applied, which amounts to just 50%. This proportion raises doubts about the claim made in the abstract that applying PCA to sliding sequence windows can maintain model performance.\n\n11.The authors state in Table 3: \"The results of Linear are adapted from the study (Zeng et al., 2023).\" However, upon reviewing the cited paper by Zeng et al. (2023), I was unable to locate the specific data presented by the authors. This discrepancy raises concerns about the reliability and accuracy of the data used in their experiments." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The idea of applying PCA within sliding windows offers a fresh perspective on dimensionality reduction for time series data. By reducing the dimensionality of the input data, the proposed method can decrease computational load, which is particularly beneficial for deep learning models dealing with large-scale or high-frequency time series data." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors propose an innovative approach for preprocessing time series data by applying Principal Component Analysis (PCA) to data within sliding windows. This method aims to extract the principal components from the data, effectively reducing dimensionality before feeding it into a deep learning network. Traditionally, it is commonly believed that applying PCA along the time dimension can disrupt temporal dependencies inherent in time series data. Contradicting this notion, the authors suggest that applying PCA to sliding sequence windows can preserve model performance while enhancing computational efficiency. They support their claims with experiments conducted on three primary tasks: time series classification, prediction, and regression." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The study exhibits several weaknesses. Firstly, it lacks clarity on dimensionality reduction along the time dimension, focusing instead on feature dimension reduction without reducing time steps, which contradicts its stated goal. Secondly, the application of PCA to the test data in classification tasks is ambiguous; applying PCA to test data is inappropriate, but without it, the claimed acceleration in inference time lacks justification. Thirdly, the study misrepresents related work by critiquing standard practices designed to prevent information leakage, and its theoretical analysis fails to support the core claim of time dimension reduction. Additionally, the experiments lack statistical validation and parameter exploration, relying on single runs with fixed principal component numbers, raising concerns about generalizability and potential overfitting. The authors make overgeneralized and absolute claims about the benefits of PCA without sufficient evidence, ignoring observed performance degradation in certain datasets. Furthermore, limited dataset diversity suggests results may be dataset-specific, and discrepancies in reported data raise doubts about reliability. Lastly, the study challenges established concepts in time series analysis without adequate empirical support, and methodological inconsistencies, such as varying the number of principal components without clear rationale, hinder reproducibility and limit the applicability of the findings." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 2 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "What makes PCA particularly effective here? Is there something unique about the space spanned by its vectors?\n\nHow does the dimensionality affect the results? A graph showing MSE versus the number of dimensions (n) would be helpful.\n\nWhen selecting the first n eigenvalues, do you choose the largest, the smallest, or select them randomly?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- It clearly introduces the problem of temporal dimensionality reduction in time series data and provides a solid rationale for using PCA.\n- The experimental setup is thorough, covering various time series tasks (classification, forecasting, and regression) and a range of model types (Linear, Transformer, CNN, RNN), which effectively illustrates the generalizability of the approach.\n- The paper strengthens its argument by presenting concrete metrics, such as GPU memory reduction and speed improvements." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper investigates using Principal Component Analysis (PCA) to reduce the temporal dimensionality of time series data in deep learning models for tasks like classification, forecasting, and regression. Traditionally, PCA has been applied to reduce variable dimensions, but this study applies PCA across time windows, aiming to maintain temporal structure while reducing redundancy and computational costs. Results show that PCA preprocessing can accelerate training and inference by up to 40% in some models, like Informer, and reduce memory usage by 30% in models like TimesNet without compromising accuracy. PCA also proves effective in noise reduction, retaining essential statistical characteristics and supporting efficient learning across different deep learning models." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The intuition behind why PCA is specifically suitable for time series dimensionality reduction is not clearly explained. Many studies have shown that using orthogonal bases (e.g., FFT, wavelets, Legendre polynomials) can improve performance and reduce dimensionality, yet the paper does not address how PCA differs or why these methods were not included in comparisons.\n- Adding comparisons with modern compression techniques, beyond linear methods and downsampling, could make the evaluation more robust.\n- Some sections, particularly on the theoretical underpinnings of PCA’s use for time series, could benefit from clearer explanations to aid reader comprehension.\n- Each table could benefit from explanations of the metrics used, clarifying what constitutes a “good” or “bad” result (e.g., lower MSE is better), which would help readers interpret the results more easily.\n- More detailed visualizations, such as diagrams showing PCA’s effects on time series structure and feature retention, could enhance clarity." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "This study employs PCA for preprocessing time series before inputting it into deep-learning models to enhance model efficiency." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024revisiting,\ntitle={Revisiting {PCA} for Time Series Reduction in Temporal Dimension},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0CvJYiOo2b},\nnote={under review}\n}" }, "abstract": { "value": "Deep learning has significantly advanced time series analysis (TSA), enabling the extraction of complex patterns for tasks like classification, forecasting, and regression. While dimensionality reduction has traditionally focused on the variable space—achieving notable success in minimizing data redundancy and computational complexity—less attention has been paid to reducing the temporal dimension. In this study, we revisit Principal Component Analysis (PCA), a classical dimensionality reduction technique, to explore its utility in temporal dimension reduction for time series data. It is generally thought that applying PCA to the temporal dimension would disrupt temporal dependencies, leading to limited exploration in this area. However, our theoretical analysis and extensive experiments demonstrate that applying PCA to sliding series windows not only maintains model performance but also enhances computational efficiency. In auto-regressive forecasting, the temporal structure is partially preserved through windowing, and PCA is applied within these windows to denoise the time series while retaining their statistical information. By preprocessing time series data with PCA, we reduce the temporal dimensionality before feeding it into TSA models such as Linear, Transformer, CNN, and RNN architectures. This approach accelerates training and inference and reduces resource consumption. Notably, PCA improves Informer training and inference speed by up to 40% and decreases GPU memory usage of TimesNet by 30%, without sacrificing model accuracy. Comparative analysis against other reduction methods further highlights the effectiveness of PCA in enhancing the efficiency of TSA models. Code is provided in the supplementary materials." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "principal component analysis (PCA)", "time series classification", "time series forecasting", "time series extrinsic regression" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/cae3a553d272ea71124861659e8713307733da09.pdf" }, "presentation": null, "primary_area": { "value": "learning on time series and dynamical systems" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/c4482cce77197e1b62a6e550dbab1ad5f101fd2a.zip" }, "title": { "value": "Revisiting PCA for Time Series Reduction in Temporal Dimension" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Some additional questions remian:\n\nCan the authors provide more details on the human evaluation process, such as annotator screening, training, and inter-annotator agreement metrics? This would help validate the human evaluation results. The paper is light on this.\nHow do you expect the methods to perform on multilingual models and non-English datasets? \nThe discussion on negative societal impacts of AI hyper-personalization (e.g. filter bubbles, opinion polarization) and is light. Authors should exand on this more. \nFinally, exploring beyond multiple choice for collecting preference data, such as open-ended text can be interesting and more useful." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "A key strength of the work is the PAPI dataset, with over 300,000 real-world subjects providing detailed responses to the IPIP-NEO-120 and IPIP-NEO-300 questionnaires. The scale is impressive.\n\nThe Personality Activation Search (PAS) method is interesting. By identifying key activation vectors that correspond to personality traits and optimally shifting activations in those directions, this approach can more effectively do personality alignment during inference. \n\n The authors also conduct a comptehensive evaluation of PAS, comparing it against prompting-based (Few-Shot, P2) and RL-based (PPO, DPO) baselines on the PAPI dataset." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes the concept of Personality Alignment for tailoring LLMs to match the preferences and behaviors of individual users or groups. The authors created a large-scale dataset called PAPI with data on behavioral preferences from over 300,000 real subjects across the Big Five personality dimensions. They also propose the Personality Activation Search (PAS) method for efficiently aligning LLMs with individual preferences during inference. PAS identifies key activation vectors corresponding to personality traits and optimally shifts activations in those directions. The authors show that PAS achieves strong performance in capturing individual preferences with high compute efficiency compared to baseline methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the PAPI dataset is impressively large, it doesnt seem to be diverse. Around 60% of subjects are female and the average age is 25 years. This skew can potentially bias the results and limit generalizability to other populations. \nAlso, PAPI dataset relies on self-report data from personality questionnaires. While this is a standard approach in personality research, self-reports can be subject to biases such as social desirability and lack of self-insight. Incorporating additional data sources, such as behavioral measures or peer ratings, could be more useful.\n\nThe evaluation of PAS focuses primarily on high-level alignment with the Big Five traits. However, personality is a complex, multifaceted construct, and individuals can vary in their expression of specific facets within each trait. \n\nThe PAPI dataset uses a multiple-choice format for collecting personality data. While this allows for structured and efficient data collection, it may limit the richness and naturalness of the responses. \nThe paper also compares PAS to prompting and RL baselines but does not include a comparison to fine-tuning the entire language model. This is an important consideration as well.S" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 5 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- In any of your experiments, did any of your tuning or evaluation methods trigger safety wall? For example, the model might refuse to answer some questionnaire questions related to consciousness or self-awareness." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "- Overall, I like this paper and I think it's a very good attempt in aligning LLMs from a personality perspective. \n\n- The paper is generally well-written and easy to read, illustrations are also made in a good quality.\n\n- It's cool to see how personality affects downstream reasoning tasks. It has always been something missing in prior personality-related LLM work. And it's definite a god step here.\n\n- The proposed method is efficient, and can provide better results compared to prompting-based methods. It may have wide applications in tailoring persona/personality-specific chatbots to end users." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces the concept of Personality Alignment for LLMs, that is, tailoring of responses to individual user preferences based on personality traits. Using the Big Five personality theory, the authors introduces the PAPI dataset modeling human personality distributions. The authors also propose a LLM tuning method based on disabling certain activation heads -- Personality Activation Search (PAS). Evaluation results demonstrate PAS’s performance and efficiency compared to traditional alignment techniques including RL- and prompting-based methods." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- Presentation: The authors claim that they collected 307k human samples to craft the PAPI dataset and the use of the IPIP-NEO-120/IPIP-NEO-300 for evaluations as part of their contributions. However, the IPIP-NEO series inventories were frequently used in prior works (e.g., Jiang et al., 2024 as mentioned in the paper), and the 307k responses are also publicly available.\n\n- The implications of the 307k PAPI dataset are not that clear. For example, in the experiments authors performed, it seems only the overall average personality tendencies and specific participants' responses are used. So what's the actual advantages of using a such large dataset?\n\n- The proposed tuning method, although efficient, but requires access to model weights. I doubt its ability to generalize such methods to black-box methods (e.g., GPTs) in the future." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Could the authors explain why the PAS method is so effective by looking at the internal hidden layer representations of the model? Intuitively, direct training approaches like DPO/PPO might yield more noticeable improvements, but the results here contradict my expectations.\n\nIn line 429, the authors mention \"Why Did Scaling Laws Fail?\" but don't seem to fully answer this question. I would like the authors to explain this from the perspective of the domain's specificity. Is it because larger models learn more general alignment, which makes it harder for them to excel at aligning with a specific personality?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The strengths of this paper can be summarized in three key points:\n\n- It contributes a new dataset (or more accurately, a dataset generation pipeline).\n- The proposed method is simple yet highly effective.\n- The experimental analysis is comprehensive, with particularly detailed descriptions of the experimental setup." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper explores the personality alignment of large language models (LLMs). Specifically, it introduces a new dataset and proposes the PAS method for personality alignment based on this dataset." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "However, I believe the paper has the following weaknesses:\n\n- The scope of the contribution is somewhat limited. I would have liked to see this method applied to a broader range of personalities, rather than being restricted to just the five personalities of the Big Five model. The authors could consider additional datasets, such as the Dark Triad or even the MBTI test (though MBTI remains controversial in psychology). Expanding in this way would enhance the paper’s overall contribution.\n\n- While the proposed method performs well on the dataset, it is actually quite similar to approaches used in many previous studies [1][2]. As a result, the novelty of the method is somewhat lacking, and the authors should be careful not to overstate their contribution.\n\n- Essentially, the proposed method is a form of personalized alignment, so the authors should compare it against more baselines [3].\n\n- Figures 6 and 7 show a significant disparity between LLM-as-a-judge evaluations and human evaluations, which makes me question the consistency and reliability of the judgment process.\n\n- Other details: In line 1304, the authors mention that human annotators come from machine learning and computer science, but ML is a part of CS. Additionally, could the authors disclose the educational background of the annotators (undergraduate or graduate)?\n\n[1] Zheng, Chujie, et al. \"On prompt-driven safeguarding for large language models.\" *Forty-first International Conference on Machine Learning*. 2024.\n\n[2] Wang, Haoran, and Kai Shu. \"Backdoor activation attack: Attack large language models using activation steering for safety-alignment.\" *arXiv preprint arXiv:2311.09433* (2023).\n\n[3] https://github.com/liyongqi2002/Awesome-Personalized-Alignment" }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We introduce Personality Alignment for language models, efficiently tailoring responses to individual user preferences, providing the‘ PAPI dataset with over 300K subjects and a practical, efficient alignment method." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024personality,\ntitle={Personality Alignment of Large Language Models},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0DZEs8NpUH},\nnote={under review}\n}" }, "abstract": { "value": "Current methods for aligning large language models (LLMs) typically aim to reflect general human values and behaviors, but they often fail to capture the unique characteristics and preferences of individual users. To address this gap, we introduce the concept of Personality Alignment. This approach tailors LLMs' responses and decisions to match the specific preferences of individual users or closely related groups. Inspired by psychometrics, we created the Personality Alignment with Personality Inventories (PAPI) dataset, which includes data from 300,000 real subjects, each providing behavioral preferences based on the Big Five Personality Factors. This dataset allows us to quantitatively evaluate the extent to which LLMs can align with each subject's behavioral patterns. Recognizing the challenges of personality alignments—such as limited personal data, diverse preferences, and scalability requirements—we developed an activation intervention optimization method. This method enhances LLMs' ability to efficiently align with individual behavioral preferences using minimal data and computational resources. Remarkably, our method, PAS, achieves superior performance while requiring only 1/5 of the optimization time compared to DPO, offering practical value for personality alignment. Our work paves the way for future AI systems to make decisions and reason in truly personality ways, enhancing the relevance and meaning of AI interactions for each user and advancing human-centered artificial intelligence." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Personality Alignment", "Large language models", "behavioral preferences of LM" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/f8f5075ac22c1f67f04cfb181360c9e65f0b6466.pdf" }, "presentation": null, "primary_area": { "value": "generative models" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/2591f5eeb5b79187828d6d7400ee27aecf8e8762.zip" }, "title": { "value": "Personality Alignment of Large Language Models" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "Please address the issues above." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "-- The paper studies an interesting problem to speed updecoding by predicting multiple tokens in parallel at higher acceptance rates than typical speculative sampling approaches.\n\n-- The proposed solution seems straightforward to implement.\n\n-- The contribution to identifying issues with existing multi-token training approaches and proposing a higher rank alternative is novel." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper borrows key idea from Gloecke et al. [1] to train multi-token predictors instead of single next word predictor. This work identifies a key flaw in [1] which is that the distributions for multiple $n$ future tokens are independent of each other thus ignoring the token interdependency. This work interprets this as a rank-1 approximation to the full distribution tensor of $n$ next tokens and proposes to improve this to a higher rank estimate. This higher rank estimate is achieved by $r$ multiple heads defining $r$ different distributions and using their mixture for the $n$-future token prediction. Training and inference method for this is discussed followed by an observation that the multi-token predictor can be used in self-speculative sampling approach where the next word prediction is made faster by using proposal distribution that predicts multiple next tokens. The experiments are mainly performed on nano-GPT model architecture trained on TinyStories dataset and also finetuning the PyCodeGPT model." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "-- The evaluation leaves a lot to be desired. Experiments are done on small datasets and small models but more concerningly, little else is provided aside from loss curves of training runs and token acceptance rates for the scheduled sampling approach. As an example, performance of these models on various benchmarks to estimate the quality of these trained models would aid in better assessment of the approach. Also, it is unclear if this approach empirically scaled to larger datasets and models effectively in terms of speed and performance.\n\n-- Comparison to other speculative sampling approaches with various draft models will give abetter idea about the improvement on speed and resources with the proposed approach.\n\n-- There is room for improvement in presentation. Figure 1 doesn't help with understanding the paper better and is confusing. Algorithm 1 can also be described more clearly. Currently, it hinges on the reader's prior understanding of speculative decoding." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "-How does the method combine with beam search?\n\n-Does the speedup increase or decrease as a function of model size?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "-Tackles an interesting and important problem\n\n-The method is written clearly. \n\n-I also find the connections to tensor decomposition interesting." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper studies multi token prediction in transformer language models. Vanilla autoregressive degressing is expensive for long outputs since it only decodes one output at a time.\n\nThe authors are inspired by the work of Gloeckle et al. 2024. In Gloeckle et al. 2024, given a context x_{t:1} the next n tokens are predicted independently (with multiple heads). As the authors point out, this amounts to a rank-1 tensor approximation of the joint probability distribution.\n\nIn this work, the authors explore higher ranks (r > 1) using CP decomposition. They draw a connection to mixture-of-experts and propose a auxiliary load balancing strategy so all the weight is not on one expert (component). \n\nThey then perform experiments validating their work." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Some confusion on experimental results: I'm a bit confused as to how much of a speed up the author's approach gives over both the approach of Goeckle et al. 2024 (i.e. rank=1) and also vanilla non-autoregressive decoding for the same level of quality. \n\nFor example in Table 1: I see that in Table 1 the final column (time per token) is not much different across all the rows?\n\nMoreover I don't quite understand Table 3.\n\nComparisons: I think the authors need additional baselines in addition to just ablations of their own approach from the related work.\n\nRelated work: The authors should also cite and discuss related work in non-autoregressive decoding (typically for neural machine translation) that has been developed for a while e.g. see below and citations therein.\n\nhttps://arxiv.org/abs/1711.02281\nhttps://arxiv.org/abs/2204.09269\nhttps://arxiv.org/abs/2012.15833" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "See above." }, "rating": { "value": 3 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "(1) The method is well-motivated and explained clearly. The connection to MoE, which motivates a load-balancing auxiliary loss, is also interesting.\n\n(2) The paper seeks to improve inference speed in large models, which is an important problem." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "One existing form of speculative decoding involves predicting k tokens at a time independently, which can be thought of as a rank-1 decomposition of the k-order joint probability tensor over those tokens. This paper instead proposes to predict the factors for a rank-r decomposition. They evaluate two instantiations of this idea: training a LM from scratch to predict this decomposition, and taking an existing LM and fine-tuning additional heads to predict this decomposition. Their experiments show that higher rank decompositions lead to higher acceptance rates in speculative decoding." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the method seems interesting and promising, the paper's experiments seem disorganized and insufficient to fully demonstrate the effectiveness of the method.\n\n(1) The majority of the results are for a 56.3M parameter trained on TinyStories, which is a very limited evaluation setting, both because the dataset is synthetic and because the setting involves retraining. There are also some experiments on head-only tuning for PyCodeGPT in Table 3, but the results in that setting are not very strong --- increasing the rank does not actually seem to actually improve inference speed for many of the models. The paper would benefit from more thorough evaluation and stronger results (especially on non-synthetic datasets, and on speeding up existing models rather than requiring retraining: for example, the evaluations done in https://arxiv.org/pdf/2211.17192 (Table 3) would improve this paper).\n\n(2) The majority of the experiments section seems to involve analysis rather than results: only tables 1 and 3 report inference times, which are the main results. I would suggest moving other plots (token acceptance rate, first token vs joint loss, etc.) to a separate analysis section.\n\n(3) There are a substantial number of issues with the experiment design that would be beneficial to address: (a) In Figure 3, it seems like hyperparameters are being selected using the test set; I would suggest using a dev set instead. (b) To make comparisons fair, I would suggest training each rank for the same amount of wall-clock time, rather than number of steps, in case higher ranks require more time per forward pass. (c) The self-speculative setup makes the results hard to interpret because each rank uses a different target model. I would suggest that each method be speculative with respect to the same target model. (d) The paper would be clearer if the experiments were described concretely: for example, the paper states that \"Our measurements were made with seq length varying from 1024\nto 4096\" (lines 408-409), but it's not clear which experiments use which sequence lengths." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1.\tIn line 113, the authors denote the input sequence as x_{t:1} and the corresponding embeddings as e_{t:1}. According to the description, the embeddings are the representations of the final transformer layer, while in Figure 1, the same value is denoted as z_t. Do z_t and e_t means the same representation, or e_t means the “input” embeddings? This notation is somewhat confusing.\n\n2.\tAre there any results on the acceptance rate for Llama 8B, not just inference time?\n\n__Typos__:\n\n1.\tIn line 116, a comma is missing before \"the conditional probabilities ...\".\n2.\tIn line 150, \"Note, that\" should be revised to \"Note that\"." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1.\tThis work identifies the limitations of recent multi-token prediction standards and proposes a more generalized approach.\n\n2.\tThe experimental results demonstrate the method's effectiveness, and the ablation study underscores the importance of the introduced components." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper focuses on speculative decoding methods that incorporate additional prediction heads. The authors conceptualize current standard approaches as rank-1 canonical tensor decomposition and propose a generalized method that extends from rank-1 to rank-r canonical tensor decomposition to approximate the joint distribution of future tokens. To enhance model training, an auxiliary loss is introduced to address weight imbalances. Experimental results highlight several key findings:\n\n1.\tIncreasing the ranks results in a decrease in joint loss.\n\n2.\tThe first token appears to have no correlation with different ranks.\n\n3.\tThe method is effective even when only the prediction heads are trained.\n\nThe proposed approach achieves notable speedups compared to autoregressive models and rank-1 baselines." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1.\tThe work lacks comparison with existing state-of-the-art methods such as Medusa, Eagle, etc., which belong to the same research domain.\n\n2.\tIn the code generation setting, the performance of averaging two accepted draft tokens is not promising.\n\n3.\tThere are several typos in this version that need revision." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "New model for multi-token prediction in transformers based on canonical probability decomposition that improves the sampling efficiency in the self-speculative decoding paradigm without compromising accuracy." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024faster,\ntitle={Faster Language Models with Better Multi-Token Prediction Using Tensor Decomposition},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0EP01yhDlg},\nnote={under review}\n}" }, "abstract": { "value": "We propose a new model for multi-token prediction in transformers, aiming to enhance sampling efficiency without compromising accuracy. Motivated by recent work that predicts the probabilities of subsequent tokens using multiple heads, we connect this approach to rank-1 canonical tensor decomposition. By generalizing it to a rank-r canonical probability decomposition, we develop an improved model that predicts multiple tokens simultaneously. This model can also be interpreted as a mixture of experts, allowing us to leverage successful techniques from that domain for efficient and robust training. Importantly, the overall overhead for training and sampling remains low. Our method demonstrates significant improvements in inference speed for both text and code generation tasks, proving particularly beneficial within the self-speculative decoding paradigm. It maintains its effectiveness across various model sizes and training epochs, highlighting its robustness and scalability." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Large language model", "Self-speculative decoding", "Multi-token prediction", "Low-rank approximation" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/6fe4ded46a26ce5b63b0b8d91cd7d350f5112c33.pdf" }, "presentation": null, "primary_area": { "value": "foundation or frontier models, including LLMs" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/8f4f0ebbb86d7ba518a01a0592a741618d66b2dd.zip" }, "title": { "value": "Faster Language Models with Better Multi-Token Prediction Using Tensor Decomposition" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "What's the connection and difference between the MSB method and the score-matching strategy (like Song et al. 2021)? What's the performance difference?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "By using time-symmetry, MSB only requires a single neural network and half of the computational expense compared to other IPFP-based algorithms." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes mirror Schrodinger bridge (MSB), a model for conditional resampling. An alternating minimization procedure is used to solve for the MSB with a theoretical guarantee. On the empirical side, the MSB method is implemented to sample from both toy distributions and image distribution from the real world." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. The theoretical results are limited to asymptotic analysis. The convergence rate is not presented;\n2. In the empirical evaluation, comparison to baseline methods is limited to the Gaussian example in Section 5.1. As a result, it's not clear how MSB compares to other methods in real-world image generation." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 4 }, "primary_area": null, "questions": { "value": "1. **Error Analysis for AMP**: The paper introduces the Alternating Minimization Procedure (AMP) in Equations (4) and (5) but lacks an error analysis. Could you elaborate on the convergence speed of AMP? Specifically, are there theoretical bounds or guarantees on the convergence rate that would enhance the theoretical foundation of your method?\n\n2. **Benchmarking Against Existing Methods**: Could you clarify which algorithms you compared with your proposed method for each example case in Section 5? Benchmarking against established methods would help situate your approach within existing literature.\n\n3. **Choice of Euler-Maruyama**: Your algorithm employs Euler-Maruyama discretization. How does this choice impact the accuracy and efficiency of solving the Schrödinger bridge problem? Have you considered alternative discretization schemes, and how do they compare in terms of performance?\n\n4. **Iterations vs. Runtime**: How does halving the number of iterations quantitatively affect running time? A breakdown of runtime reductions relative to iteration count would provide a clearer picture of the efficiency gains.\n\n5. **Quantitative Metrics and Proximity Definition**:\n - **Metrics for Resampling Quality**: The paper lacks specific metrics to assess resampling quality in each example case. What metrics do you use to evaluate how well the method preserves the integrity of the original distribution during resampling?\n - **Proximity Definition**: An empirical definition of proximity would clarify how closely generated samples align with input data. For example, on MNIST, how would a proximity value of a digit-5 image with \\(\\sigma=1\\) compare to that of a digit-3 image with a different \\(\\sigma\\)?\n\n6. **Experimental Setup Details**: Although the paper claims empirical validation across various domains, the experimental setup lacks specificity. Could you provide more details on the datasets, experimental procedures, and metrics used to assess performance?\n\n7. **Quantitative Results and Comparisons**: Could you include performance metrics and comparisons that demonstrate the robustness and potential advantages of mirror Schrödinger bridges over existing techniques?\n\n8. **Scalability to High-Dimensional Data**:\n - **Performance on High-Dimensional Benchmarks**: The scalability of your method to high-dimensional data is not discussed. Could you provide insights into its performance and computational complexity on high-dimensional or large-scale datasets?\n - **Strategies for High-Dimensional Data**: What strategies do you propose to ensure the scalability and efficiency of mirror Schrödinger bridges when applied to high-dimensional data or data on manifolds? Are there modifications or optimizations that could improve performance in these scenarios?\n\n9. **Interpretation of Figure 2**: In Figure 2, the resampling estimate with $\\sigma=1$ appears to produce more concentrated samples compared to the original samples. Could you explain this behavior? Is there some form of geometric or manifold enhancement occurring in the resampling process?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 4 }, "strengths": { "value": "The paper introduces the mirror Schrödinger bridge framework, an approach that differs from traditional Schrödinger bridges by focusing on mapping a distribution onto itself rather than between distinct distributions. This self-mapping approach directly addresses the challenge of conditional resampling within the same distribution, opening up new possibilities for generating in-distribution variations of data points. By incorporating time symmetry and the Alternating Minimization Procedure (AMP) to establish theoretical foundations, the paper presents an innovative solution to resampling. The algorithm also leverages time symmetry to train a single neural network for modeling the diffusion process drift, enhancing computational efficiency.\n\nThe paper includes solid theoretical foundations and provides comprehensive convergence proofs in the total variation metric, even in infinite-dimensional state spaces. The AMP is carefully developed and shown to converge to the mirror Schrödinger bridge, ensuring methodological consistency. The algorithm’s implementation is efficient, theoretically reducing computational overhead by half compared to Iterative Proportional Fitting Procedure (IPFP)-based methods. Empirical evaluations across several applications support some theoretical claims, demonstrating the method’s capability to generate high-quality proximal samples for tasks like data augmentation and generative modeling. The organized presentation of theoretical and empirical findings underscores the paper’s contribution to the field.\n\nAdditionally, the paper is well-written and structured, guiding the reader through complex concepts with clarity. The transition from theoretical foundations to practical algorithmic details is seamless, ensuring a coherent flow. Most definitions and problem formulations are clearly presented. Explanations of AMP and iterative schemes enhance understanding, while algorithmic pseudocode supports practical comprehension." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces the mirror Schrödinger bridge, a method for addressing the resampling problem when the initial and target distributions are identical. Unlike traditional Schrödinger bridges, which are designed for mapping between two distinct distributions, the mirror Schrödinger bridge is formulated specifically for self-mapping within a single distribution. This unique approach facilitates the generation of in-distribution variations of data points, allowing for conditional resampling that maintains the original distribution’s integrity.\n\nThe authors develop a theoretical foundation for this method, employing time symmetry and the Alternating Minimization Procedure to establish convergence in the total variation metric, even for infinite-dimensional state spaces. This achievement addresses the challenging issue of convergence in high-dimensional settings. Additionally, the algorithm capitalizes on the time symmetry inherent in the problem, enabling it to model the diffusion drift with a single neural network. This innovation significantly reduces computational costs, effectively halving the effort compared to Iterative Proportional Fitting Procedure based approaches.\n\nEmpirical evaluations underscore the practical value of the mirror Schrödinger bridge across diverse applications, highlighting its capability to produce high-quality proximal samples that are valuable for tasks like data augmentation and generative modeling. In summary, this research claims to provide a theoretically rigorous and computationally efficient solution for conditional resampling within the same distribution, combining solid theoretical contributions with practical algorithmic advancements." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the paper introduces mirror Schrödinger bridges, the framework retains substantial mathematical similarities to traditional Schrödinger bridges. Could you clarify the key mathematical distinctions between the two approaches? The theoretical analysis would benefit from an error analysis for the Alternating Minimization Procedure (AMP) outlined in equations (4) and (5), which would provide valuable insights into the convergence rate of the proposed approach. The absence of such an analysis limits our understanding of the efficiency and accuracy of the AMP. \n\n\n\nOn the practical side, the algorithm does not include comparisons with other established approaches in the literature, nor is there any discussion regarding the impact or choice of specific discretization methods, such as the Euler-Maruyama scheme, in addressing this problem. Furthermore, the paper lacks a detailed explanation of all the Figures, which hinders the reader's ability to assess how the proposed method performs relative to existing methods and to understand the extent to which halving the iterations reduces runtime quantitatively. \n\nThe paper also lacks a quantitative flow for evaluating the equality of resampling across examples and fails to specify metrics used to assess performance. An empirical definition of \"proximity\" would add clarity. For instance, it is unclear whether a proximity value of 5 remains constant with $\\sigma=1$ or changes to 3 with a different $\\sigma$, and the way proximity is quantified in empirical examples remains vague. Although the paper claims validation across various application domains and provides some information on the experimental setup and datasets, it lacks sufficient detail on the specific metrics used for performance assessment, making it difficult to evaluate the method's effectiveness and robustness relative to existing techniques. \n\nFinally, while the paper emphasizes algorithmic simplifications that reduce computational costs by training only one neural network, it does not address the scalability of the method to high-dimensional data. The absence of any analysis on how the method performs as the data dimensionality increases leaves questions about its applicability to high-dimensional settings, which are increasingly relevant in practical applications." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "N/A" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Comments:\n- Line 053: maybe write what \"\\delta\"-measure means (or just say Dirac measure).\n- The paragraph just above figure 3 is very unclear. I'm not sure what the rows and columns are meant to refer to here..\n- The ability to use one neural network is not surprising from the connection between EOT and the SB problem. Equation (24-25) in the work by Feydy et al. (2019) precisely proposes some kind of fixed-point equation on one potential function (whereas for entropic OT between two measures, there are typically two potentials to optimize over via the Sinkhorn algorithm)\n- Question: Is there any hope to provide a *rule* for choosing the amount of noise added in the process of generating images?\n- Question: The choice of OU process appears entirely arbitrary. Why not consider standard Brownian motion as the reference process?" }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "This paper proposes to alleviate some computation burden of training SBs by proposing a learning algorithm that learns the SB between a measure and itself. Still faithful to the generative modeling paradigm, they learn the SB from a data distribution to itself, with the goal of starting at existing samples and generating diverse ones. The writing is relatively clear, and Figure 2 is especially clear at describing the phenomenon of \"starting from an existing sample\" and, given enough noise, learns to go somewhere else in the distribution." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This work studies the Schrödinger bridge (SB) between a distribution and itself, unlike most of the literature which focuses on the SB between two distributions (e.g., the standard Gaussian, and the data distribution). The authors propose this model as a means for conditional resampling of a distribution, where the \"noise\" induced by the bridge process allows them to obtain new samples which are in-distribution. They demonstrate their approach on many experiments, and have some proofs of their technical results." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "While the computational burden of having to train two neural networks instead of one is appealing, there is little comparison between MSB and DSB or DSBM in terms of quantifying sample versatility (they performed some comparisons in the Gaussian case, but these are far from conclusive). For instance, in Figure 3: is there a certain $\\sigma_\\star$ after which the data generated by the MSB changes classes? This is likely hard to prove theoretically, but knowing if there exists some threshold after which data stops being the same class would be very interesting." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "When training, how are the data split? Is there any pairing done between the data points? How would that affect the mapping between the two sets. \n\nAre there other mean reverting processes that you can consider other than OU processes? How do these affect the performance of the method?\n\nIs there a set of results one can consider that consist of analyzing the regularity of the path measures?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The authors provide an interesting perspective on the Schrödinger bridge problem and provide a new technique for fitting a path measure connecting an initial condition given by itself to itself.\n\nThe method is fairly straightforward to implement and say to analyze.\n\nThe method provides optimality with respect to relative entropy, which is a nice property of the sample paths." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The authors consider a modification to the original Schrödinger bridge problem where they consider the marginals to be the empirical measure of the data. This learns a coupling between two sets of the same data. The coupling is designed to optimal in the relative entropy sense. They modify the iterative proportional fitting procedure such that they project in the direction that minimizes the KL divergence in one step and then in the reverse KL divergence in the next step due to the analytical feasibility of the step. They propose a practical algorithm for computing the projections using a change of measure technique and optimizing the drifts." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "My main concerns come with the empirical evaluation of the method. \n\nWhile the method has nice motivation, empirically the results do not seem to be impressive. In the cases of low \\sigma, the variation is very small compared to the initial condition. This is likely because the method is estimating a map between itself that is effectively an OU process. Furthermore, when the $\\sigma$ is large, the methods appear to be much more corrupt and lose some of the important features. \n\nIn general the performance of the method does not seem to be well studied. Since one of the motivations the authors mentioned was based on the path measure being optimal with respect to relative entropy, I would show some of these results on the regularity of the path space." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "Conditional resampling by solving the Schrödinger bridge problem between a distribution and itself" }, "_bibtex": { "value": "@inproceedings{\nanonymous2024through,\ntitle={Through the Looking Glass: Mirror Schr\\\"odinger Bridges},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0F1rIKppTf},\nnote={under review}\n}" }, "abstract": { "value": "Resampling from a target measure whose density is unknown is a fundamental problem in mathematical statistics and machine learning. A setting that dominates the machine learning literature consists of learning a map from an easy-to-sample prior, such as the Gaussian distribution, to a target measure. Under this model, samples from the prior are pushed forward to generate a new sample on the target measure, which is often difficult to sample from directly. In this paper, we propose a new model for conditional resampling called mirror Schrödinger bridges. Our key observation is that solving the Schrödinger bridge problem between a distribution and itself provides a natural way to produce new samples from conditional distributions, giving in-distribution variations of an input data point. We show how to efficiently solve this largely overlooked version of the Schrödinger bridge problem. We prove that our proposed method leads to significant algorithmic simplifications over existing alternatives, in addition to providing control over conditioning. Empirically, we demonstrate how these benefits can be leveraged to produce proximal samples in a number of application domains." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "entropic optimal transport", "schrödinger bridge", "stochastic differential equations", "sampling" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/6283ff95f5da61abd862bf58ddf31ffd31ab2236.pdf" }, "presentation": null, "primary_area": { "value": "probabilistic methods (Bayesian methods, variational inference, sampling, UQ, etc.)" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": null, "title": { "value": "Through the Looking Glass: Mirror Schrödinger Bridges" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Are the OOD detector and noise management incorporated with other data augmentation baselines, e.g., SynthER and ATraDiff?\n\n2.In Section 6, a very specific environment is adopted to show the advantages of reverse synthesis over normal synthesis, what is the reason/possible explanation that normal synthesis with OOD detector even produces significantly more In2Out than normal synthesis without OOD detector? (18.2%11.2)\n\n3. For other questions, please questions raised in Weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The ideas in this paper are interesting and novel, where to my knowledge, this is the first work that utilizes the concept of generating the reverse trajectories to address the distribution shift issue. The paper is clearly written and well-motivated. The effectiveness of the proposed method is verified in various environments, and ablation studies are also conducted to validate the effectiveness of different components of the proposed method." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "To address the distribution shift issue in offline reinforcement learning, this paper proposes a novel diffusion-based data augmentation technique, namely RTDiff, where rather than generating forward trajectories, the reverse one is synthesized, which is also the paper's main contribution. Furthermore, the performance of RTDiff is enhanced with the introduction of trajectory length control and noise management. Experimental results show the effectiveness of the proposed approach." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "The main concern of the paper is whether the reverse synthesis can actually address the issue of distribution shift. As it is not state clearly whether the OOD detector is incorporated with other data augmentation baselines, e.g., SynthER and ATraDiff, it is not very certain whether the better performance achievement of RTDiff is due to the reverse synthesis or the using of the OOD detector. In Section 6, an analysis is conduct by using a simple illustrative environment to show why reverse synthesis avoids issues present in normal synthesis. However, on the one hand, it is better to directly use an environment used in the experiments for the analysis, on the other hand, it is confusing why the reverse synthesis generates trajectories that move from the dangerous area to the upper or lower areas while normal synthesis generates trajectories that start from the lower area and enter the middle dangerous area. Do these two approaches both start from a state in the offline data, and then generate the trajectories in different ways? If the OOD detector is used in the normal synthesis, can the dangerous areas also be avoided? Moreover, the theoretical contribution of the proposed is not very significant. \nIf the concerns can be addressed, I would like to raise my score." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. Can the authors discuss in detail the differences between this paper and previous similar works (like ROMI)?\n2. This method is built on IQL, TD3BC, and CQL. Have there been any adjustments to the hyperparameters of these methods after using data augmentation?\n3. Is it possible to conduct a quantitative evaluation of the quality of the generated trajectories? For example, assessing the model error of the generated trajectories, etc." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. The article is well-written, with clear expression and logic, effectively reflecting the main argument.\n2. The experiments are very comprehensive. The authors validate the effectiveness of their method across a series of tasks, including both proprioceptive observations and visual observations." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper proposes RTDiff, a novel diffusion-based data augmentation technique that synthesizes trajectories in a reverse direction. Such reverse generation naturally mitigates the risk of overestimation by ensuring that the agent avoids planning through unknown states. RTDiff also introduces some other tricks including flexible trajectory control and noise management to improve sythesis quality. Emprirical results show the advantage of reverse generation over forward generation." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. I believe the authors miss a key related work (called ROMI [1]), which is the first to propose using reverse trajectory generation in the field of offline reinforcement learning. The motivation described for reverse trajectory generation in these two work is also very similar. Therefore, considering that this paper is neither the first to propose the use of reverse trajectory generation in offline RL nor the first to use diffusion for data augmentation, I would say the novelty of this paper is quite limited.\n2. I think this paper lacks comparisons with model-based offline RL methods in the experimental section. According to my understanding, using a diffusion model for data augmentation essentially falls under the same category as previous model-based offline RL methods. For instance, MOPO [2] essentially generates a batch of synthetic samples to supplement the original samples. Therefore, the authors should compare some model-based offline RL methods, such as MOPO, RAMBO [3], etc.\n\n[1] Wang et al. \"Offline Reinforcement Learning with Reverse Model-based Imagination\" (NeurIPS'21)\n[2] Yu et al. \"MOPO: Model-based Offline Policy Optimization\" (NeurIPS'20)\n[3] Rigter et al. \"RAMBO-RL: Robust Adversarial Model-Based Offline Reinforcement Learning\" (NeurIPS'22)" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "See weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "1. This paper is well-written and easy to follow.\n2. RTDiff introduces reverse trajectory generation, an OOD detector, and a noisy control method to achieve efficient and reliable sample generation. These approaches are intuitively sound.\n3. The authors conduct extensive benchmark and ablation experiments. Experimental results demonstrate that RTDiff outperforms previous baselines on both vector-based and pixel-based tasks, and each component—reverse trajectory generation, the OOD detector, and noisy control—exhibits its effectiveness." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces a novel diffusion-based data augmentation method, RTDiff, for offline reinforcement learning. First, RTDiff mitigates the data distribution shift issue present in previous data augmentation methods by generating reverse trajectories instead of forward ones. Second, RTDiff trains an out-of-distribution (OOD) detector to truncate the OOD segments of generated trajectories, further enhancing sample authenticity. Finally, the authors propose a new noisy control method to improve sample generation efficiency. Experimental results validate the effectiveness and efficiency of RTDiff and different components across both vector-based and pixel-based tasks." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. Reverse trajectory generation is not new to offline RL, and the paper lacks a discussion and experimental comparison of prior works, such as [1].\n2. The paper lacks a clear and well-reasoned explanation of the issues with previous data augmentation methods. In lines 88-90, this paper claims that previous data augmentation methods suffer from the data distribution shift, potentially leading to value overestimation. However, since all of these works use offline RL algorithms, I think data distribution shift is not the key issue. On the contrary, data augmentation is only effective when the generative model can produce samples that differ from the training data.\n3. Generated data fidelity is a more critical factor. The paper lacks a quantitative evaluation of the fidelity of generated samples.\n\n[1] Offline Reinforcement Learning with Reverse Model-based Imagination. NeurIPS, 2021." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "Please see weaknesses." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "The reverse generation method, which has been empirically verified, effectively reduces the augmentation of data in risky regions. This concept is intuitively illustrated in the diagram." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "Traditional offline reinforcement learning methods often introduce conservative biases to limit exploration to familiar regions, but this can restrict an agent's ability to generalize. While recent approaches use generative models to expand offline datasets, they can overestimate synthesized data, particularly when it includes out-of-distribution samples. RTDiff is introduced to address this, which is a diffusion-based data augmentation technique that creates trajectories in reverse, moving from unknown to known states. This reverse approach reduces the risk of overestimation by ensuring the agent avoids planning through unfamiliar regions. It also supports generating longer trajectories, utilizing diffusion models effectively while maintaining reliability. RTDiff further optimizes the process with flexible trajectory length control and noise management to improve generation efficiency." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1.Previous research [1] introduced a reverse data generation approach using a transition model. In the current work, the vanilla model is replaced with a diffusion model, yet the fundamental concept remains unchanged, limiting the overall contribution.\n\n2.The explanation relies on an intuitive diagram, but it would be more effective to demonstrate several specific cases. Identifying which states are risky. For example, some states are prone to be overestimated and easily generated by a forward model, but the reverse model effectively avoids generating them.\n\n3.Minor errors are present, such as in line 156, where \"dat\" should be corrected to \"data.\"\n\n4.Environments based on visual data should be included in the analysis.\n\n[1] Wang, Jianhao, et al. Offline reinforcement learning with reverse model-based imagination." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": { "value": "@inproceedings{\nanonymous2024rtdiff,\ntitle={{RTD}iff: Reverse Trajectory Synthesis via Diffusion for Offline Reinforcement Learning},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0FK6tzqV76},\nnote={under review}\n}" }, "abstract": { "value": "In offline reinforcement learning (RL), managing the distribution shift between the learned policy and the static offline dataset is a persistent challenge that can result in overestimated values and suboptimal policies. Traditional offline RL methods address this by introducing conservative biases that limit exploration to well-understood regions, but they often overly restrict the agent's generalization capabilities. Recent work has sought to generate trajectories using generative models to augment the offline dataset, yet these methods still struggle with overestimating synthesized data, especially when out-of-distribution samples are produced. To overcome this issue, we propose RTDiff, a novel diffusion-based data augmentation technique that synthesizes trajectories *in reverse*, moving from unknown to known states. Such reverse generation naturally mitigates the risk of overestimation by ensuring that the agent avoids planning through unknown states. Additionally, reverse trajectory synthesis allows us to generate longer, more informative trajectories that take full advantage of diffusion models' generative strengths while ensuring reliability. We further enhance RTDiff by introducing flexible trajectory length control and improving the efficiency of the generation process through noise management. Our empirical results show that RTDiff significantly improves the performance of several state-of-the-art offline RL algorithms across diverse environments, achieving consistent and superior results by effectively overcoming distribution shift." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "Reinforcement Learning", "Diffusion Model", "Reverse Synthesize" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/d2871f8fbecca7332011d381040dfe22407a2f63.pdf" }, "presentation": null, "primary_area": { "value": "reinforcement learning" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/15a8e35d680b784653f8ee42d8199420a0277013.pdf" }, "title": { "value": "RTDiff: Reverse Trajectory Synthesis via Diffusion for Offline Reinforcement Learning" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]

[ { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "- Does s=1 mean that the model is trained as a single variational autoencoder? Or is this truncation only used for sampling? (paragraph about the training is not clear to me)\n\n- naive question for my understanding: At the beginning of Section 5.1, the authors say that they consider time series of flow field from j= 1 to T. It is not clear how the time series are handled here. Could the authors clarify this point?\n\n- How do the authors handle the high fluctuations areas of the domain? It seems that some region of the domain have a highly turbulence flow field (low pressure vortex) and this would require a more flexible model in this specific area. \n\n- Did the authors try to change the value of the initial input $x_{init}$? \n\nI look forward to reading the answers of the authors and I can change my score depending on their answers." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "The paper reads very well and provides some good contributions to the development of diffusion models. \n\n - Although the presented approach is based on some known results (Tweetie's formula, ancestral sampling), the proposed solution seems very efficient in practice and leads to lower computational costs. Both idea's exploit the denoising Tweedie's formula in forward and backward sampling.\n\n - Including the truncation in the training is smart trick that eases the training. \n\n - The numerical experiments show that the presented approaches achieve better performances than traditional DDPM sampling and reduce the computational costs.\n\n- The approach seem to be very efficient in physic based simulation of flow field. I think this is a very good contribution in this specific domain." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper introduces a truncation approach and an iterative refinement process in the sampling procedure of the Denoising Diffusion Probabilistic Model (DDPM) that enable to reduce the number of function evaluations without decreasing the accuracy. The first method proposes to stop the sampling process at an earlier time point and to estimate the denoised sample using Tweedie's formula. The second method uses the forward diffusion for a given shorter noise schedule and the denoised sample is approximated using Tweedie's formula. The authors show the efficiency of the approach on the simulations of airflow field." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Although the approach seems efficient in the presented numerical experiments, there is a number of points that need to be clarified. \n\n- Tweedie's formula is well known. It seems that this has been already used in some previous works such of [Delbracio et Milanfar, 2024]. In their work, the authors provide some intermediate reconstructions through this formula. They show that by adding some stochastic steps \nthey can get better performances than state-of-the-art. I know the model is not the same but it would have been interesting to highlight the links with this work because they seem very closely related. May the authors comment on that point?\n\n- The results show that TS outperforms traditional DDPM by using s=1. This means that on this specific problem, there is no need to sample intermediate diffusion steps. I wonder whether this aspect is problem specific or this happens for a wider range of problems. Is this result expected? \n\n- This would have been interesting to see how this method perform against traditional DDPM sampling on image reconstruction. Indeed, DDPMs usually perform well on such problems.\n\n- The training data are deterministic sequences of flow field data. It would have been interesting to observe how the model performs on noisy dataset. \n\n- Overall, I like the idea of truncation and iterative refinement, but since there is no major theoretical contribution in this work, I would have liked to see more numerical results. The claim \"Truncation is all you need\" would have been justified if the authors had included numerical results on different applications. So far, the paper makes very good contribution in this specific domain, and proposes an interesting approach to reduce computational costs.\n\nDelbracio et Milanfar, 2024, Inversion by Direct Iteration: An Alternative to Denoising Diffusion for Image Restoration. TMLR." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "None. This is original work and there are no ethical issues" }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "The experimental results are very good, and I hope to add more theoretical analysis. I will be happy to improve my score in subsequent discussions." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "Advantages:\n\n1. As we all know, if we consider an infinite boundary heat equation (diffusion process), or consider long-term diffusion, then the usefulness of a long part of the noise addition/noise reduction behavior is not that great. The article fully considers and utilizes the diffusion behavior in a finite time (truncation), thus achieving a balance between accuracy and running time, which is a very good point.\n\n2. Truncated sampling reduces the uncertainty of the sample to a certain extent, or increases the accuracy of the sample response distribution.\n\n3. Based on the content of the appendix(in particular, part D), the experimental effect is very significant. In other words, iterative refinement even makes up for the truncated sampling to a certain extent (the useful information and samples that are truncated)." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This article first analyzes DDPM and finds that after an appropriate truncation (stop diffusion), the model has high fidelity and high-efficiency sampling performance. On this basis, an iterative refinement method is introduced to further improve accuracy and long-term stability." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "Disadvantages:\n\n1. I want to know whether there is a mathematical inference for the truncation sampling standard, or whether it is completely based on experience, that is, truncation and retention (importance) interpretability.\n\n2. The purpose of refinement iteration and truncation sampling is to improve efficiency while ensuring a certain degree of accuracy. I think this requires a game. How to achieve such a balance? Is there a more rigorous mathematical explanation?\n\n3. I think experiments can increase the breadth. One is to compare with a more general SDE instead of just with DDPM (and Kohl's 2024 work). In addition, for experiments, do you consider more general PDE solutions?" }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 4 }, "contribution": { "value": 3 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "**Q1:** Given that PDE-refiner is conceptually relatively similar to iterative refinement, I am surprised to see a quite large performance difference between the two methods in Appendix D.1. Can the authors explain the reasons behind this large performance gap?\n\n**Q2:** The truncation of the last steps of the reverse diffusion process seems to be equivalent to a modification of the noising schedule: we can choose the noise schedule $\\beta_t$ such that the first step in the forward diffusion process has already a quite low signal-to-noise ratio (in line with the level corresponding to the last step before truncation in the reverse process), and afterwards noise is added gradually as per usual, while reducing the total amount of steps in the forward process in line with the skip percentage. Can the authors provide their thoughts on this perspective and whether or not they agree? If they agree, can they comment on why the noise schedule that is equivalent to the truncated process is a good choice for this problem setting relative to other problem settings, and in this way place their contribution in the broader context of noise schedules?\n\n**Q3:** Please comment on W1-W3." }, "rating": { "value": 6 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "**S1:** The experimental setup is rigorous, comparing methods in both pointwise metrics and relevant physics-based metrics to provide complementary perspectives on their performance for three relevant datasets. Moreover, many additional results and baselines can be found in the appendix, making an extensive empirical evaluation overall.\n\n**S2:** Two methodological contributions are evaluated: truncation of the last steps of the reverse diffusion process, and iterative refinement, which optimizes the inference sampling schedule such that less denoising steps are required. Both contributions are aimed at reducing the number of function evaluations to improve computational complexity of diffusion models for physics simulations. This is a relevant research direction, since reducing computational complexity of computational procedures is one of the primary use-cases of neural simulation models, for which diffusion is emerging as a promising modeling approach. \n\n**S3:** The paper is well-written, and the explanations of the proposed algorithms are intuitive and easy to follow and understand. The clear structure of the text helps the reader to efficiently navigate the paper." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper studies the application of diffusion models to physics simulations. Over the past years, neural networks have emerged as surrogate modeling approach for physics simulations, with a key use-case being computationally efficient inference. However, for this purpose, diffusion models have the drawback of requiring many function evaluations due to their iterative ancestral sampling procedure. To this end, the authors propose two contributions: (1) truncation of the last steps of the reverse diffusion process, and (2) iterative refinement, which considers a much shorter noise schedule at inference time. Both methods reduce the number of function evaluations and thereby increase sampling speeds. Moreover, the empirical results demonstrate that accuracy is generally maintained and sometimes even improved compared to standard expensive sampling procedures." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "**W1:** While reading the text, I found it difficult to distill what the key differences between iterative refinement and PDE refiner are (Lippe et al., 2023). Does it have to do with the greedy optimization method of the refinement schedule (to my knowledge PDE refiner uses a fixed schedule), or details in the formulation of the diffusion process (a nonzero vs zero drift term in IR and PDE refiner respectively), or something else? Since both IR and PDE refiner are quite similar, it would be good if the ‘method novelty’ paragraph explicitly contrasts the two approaches and highlights their differences. Additionally, if the greedy optimization of $\\gamma$ is a core novelty relative to PDE refiner, then it would be beneficial to explain this more elaborately in the main text rather than the appendix, since it would be a key aspect of one of the contributions in this case.\n\n**W2:** One of the goals of the paper is to show that the proposed methods close the gap between the diffusion models and deterministic baselines. However, most of the results in the main text (both tables and plots) focus only on diffusion models. It would be relatively straightforward to also show the results of one or two deterministic methods that are considered by the authors in part of the plots and tables, for example in Figure 2 and Table 2. This would help the reader to get a better understanding of the tradeoffs of existing diffusion-based approaches, deterministic approaches, and the proposed methods without taking additional space in the paper.\n\n**W3:** The conditioning on the autoregressive step size (j in Sec. 2 of the paper) is already introduced in Gupta et al (2022), and as such cannot be claimed as a contribution of the paper (currently point 1 of the contributions listed in the introduction). Since this is not a core point in the rest of the paper, it seems that this can straightforwardly be removed from the list of contributions in the introduction without affecting the rest of the work and the core contributions significantly.\n\n**References:**\n\nGupta, J. K., & Brandstetter, J. (2022). Towards multi-spatiotemporal-scale generalized pde modeling. arXiv preprint arXiv:2209.15616.\n\nLippe, P., Veeling, B., Perdikaris, P., Turner, R., & Brandstetter, J. (2023). Pde-refiner: Achieving accurate long rollouts with neural pde solvers. Advances in Neural Information Processing Systems, 36." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": { "value": "See above." }, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 2 }, "primary_area": null, "questions": { "value": "- Could you theoretically or intuitively clarify the effectiveness of the pre-trained diffusion model in the IR sampling procedure at noise level $t = \\gamma $? The distribution $x_{init}$ at noise level $t = \\gamma $ appears to differ from the distribution $x_0$ at the same level $t = \\gamma $. Additionally, in Equation 6, is it ensured that the error between the final output and the $x_0$ remains sufficiently small, such that $E[‖x _0^N - x_0 ‖_2 ]<\\epsilon $?\n\n- Could you also clarify (experimental support would be helpful) line 288 (2), which states that IR sampling does not require data consistency to $\\hat{x}_0$? I mean why the proposed method does not require the data-consistency? Enforcing data consistency to $\\hat{x}_0$ could be plugged in after line 7 in Algorithm 2 to improve accuracy in a single iteration, without compromising the number of iterations needed. Also, related to the above question, data consistency could help reduce the error towards zero, such as [1] and [2]. \n\n[1] A physics-informed diffusion model for high-fidelity flow field reconstruction, 2023.\n\n[2] Diffusionpde: Generative pde-solving under partial observation, 2024." }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 3 }, "strengths": { "value": "- Interesting topics – Generative diffusion models as surrogate model for fluid simulations.\n- Clear motivation – Reduce computation costs in the generative process using a novel reverse sampling approach." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "This paper introduces Truncated Sampling Models (TSMs) and Iterative Refinement (IR) to improve the efficiency and fidelity of Denoising Diffusion Probabilistic Models (DDPMs) for fluid simulations by enabling reduced steps sampling through truncation of the diffusion process. These methods significantly reduce inference time and improve stability over long rollout horizons for turbulent flow and airfoil simulations." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "- The paper lacks clarification of superiority of the proposed methods over other surrogate models, such as neural operators which are currently the most widely used ML-based surrogate models in fluid dynamics for speed and accuracy. Given that a primary contribution of this paper is reducing time costs, a more thorough comparison with advanced neural operators – either highlighting the proposed method’s improved accuracy at similar time costs or its time efficiency at comparable accuracy – would strengthen the argument. However, there are few comparisons with neural operators in the experiments in main text; although Unet is included, more advanced neural operators should be included. Additionally, the proposed method does not appear to clearly outperform Unet.\n- The title may mislead some readers, as “physics-based simulations” implies a broad range of applications, while the paper is mostly on fluid dynamics. To improve clarity, I recommend replacing “physics-based simulations” to “Fluid Dynamics Simulations”. Alternatively, the authors could clarify whether they intend to generalize their approach to other physics-based simulations beyond fluid dynamics or provide examples of how the method could be applicable to other physics domains." }, "withdrawal_confirmation": null }, { "TLDR": null, "_bibtex": null, "abstract": null, "anonymous_url": null, "authorids": null, "authors": null, "code_of_conduct": { "value": "Yes" }, "code_of_ethics": null, "comment": null, "confidence": { "value": 3 }, "contribution": { "value": 2 }, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": { "value": [ "No ethics review needed." ] }, "keywords": null, "large_language_models": null, "no_acknowledgement_section": null, "other_comments_on_LLMs": null, "paperhash": null, "pdf": null, "presentation": { "value": 3 }, "primary_area": null, "questions": { "value": "1. **Regarding W1** - I’d be curious to see how the methods perform on dynamical systems with more complicated nonlinear dynamics, for example the Kuramoto-Sivashinsky equation used in PDE-Refiner (Lippe et al. [3]). That is a fourth-order non-linear equation where correctly capturing the high frequencies seems to be more important than in other settings, so it would be interesting to see how you compare there with benchmarks such as PDE-Refiner [3] (which can also be interpreted as a refinement of an MSE-trained one-step prediction). \n2. **Regarding W2** - Do you have any intuition why these methods outperform DDPM and whether the DDPM baseline could be improved?\n3. **Regarding W4** - Could you provide some experiments that test the long rollout performance of these sampling methods vs. traditional ones? Including frequency spectra of generated states would also help.\n4. **Regarding Minor 2** - It seems that in general the optimum $j$ for these methods is between [2, 4], but have you noticed any significant patterns? Were there differences between interpolation and extrapolation tasks?\n5. **Eq (5).** I am not sure I understand this equation. I will omit the $\\theta$ in the $p_{\\theta}$ subscript because the equations do not render correctly. If $p^T(x_T, \\mathbf{x}_0, j) = \\mathbf{x}(j \\cdot \\delta t)$, then wouldn’t $\\mathbf{x}(2j \\cdot \\delta t) = p^T(x_T, \\mathbf{x}(j \\cdot \\delta t), j) = p^T(x_T, p^T(x_T, \\mathbf{x}_0, j), j)$ (i.e., we still start from white noise $x_T$, but we now condition on the output of the previous step)? In my mind $\\mathbf{x}(\\tau_f) = p^T(x_T, p^T(...p^T(x_T, \\mathbf{x}_0, j)...), j)$, but maybe I didn’t interpret the equation correctly.\n6. When comparing the results in Table 5 vs. Tables 6 and 7 - Shouldn’t the metrics corresponding to, for example, Fturb TSM T80 s =0.75 ($j=2$) from Table 5 ($3.63 \\pm 1.95$) be the same as in TSM T80, s=0.75 with optimal $j=2$ ($3.43 \\pm 3.03$) Table 7? What’s the difference between these settings?" }, "rating": { "value": 5 }, "reciprocal_reviewing": null, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": { "value": 2 }, "strengths": { "value": "1. **Relevant topic.** Lately, there has been increasing interest in modelling dynamical systems with diffusion models due to their probabilistic nature. However, many works (Kohl et al. [1], Shysheya et al. [2], Lippe et al. [3]) acknowledge that diffusion models tend to be computationally costly, and techniques that reduce this cost would be greatly beneficial. This is exactly the problem this paper aims to address.\n2. **Clear distinction from other works.** The paper clearly delineates its contributions from the already existing work, and how the proposed sampling techniques differ from other approaches.\n3. **Good experimental evidence.** The paper provides good empirical evidence, with experiments on three diverse datasets, and using a wide range of metrics.\n4. **Well-structured, clear writing.** Overall, I found the structure and writing clear.\n\n[1] Kohl, G., Chen, L., & Thuerey, N. (2023). Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation.\n\n[2] Shysheya, A., Diaconu, C., Bergamin, F., Perdikaris, P., Hern'andez-Lobato, J.M., Turner, R.E., & Mathieu, E. (2024). On conditional diffusion models for PDE simulations.\n\n[3] Lippe, P., Veeling, B.S., Perdikaris, P., Turner, R.E., & Brandstetter, J. (2023). PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solvers. ArXiv, abs/2308.05732." }, "student_author": null, "submission_guidelines": null, "summary": { "value": "The paper addresses the high computational requirements of existing diffusion-based techniques for modelling dynamical systems. They propose two sampling techniques that lead to good sample quality with only a few NFEs. The first one requires modifications to the training process and is performed by truncating the diffusion process close to the clean data. The second one is compatible with pre-trained DDPMs and proposes an iterative refinement based on Tweedie’s formula. The paper provides extensive experimental evidence on three datasets: incompressible and compressible turbulent flow (2D) and airfoil flow simulation (3D)." }, "supplementary_material": null, "title": null, "venue": null, "venueid": null, "weaknesses": { "value": "1. **Unclear how the method generalises to other settings.** I acknowledge that the aim of the paper is to study the efficacy of the proposed sampling methods in the context of dynamical systems. As the authors mention, the fact that they lead to good results is probably due to:\n- The characteristics of the data distribution—states with fairly coarse resolutions, where the high frequency information has been lost during downsampling.\n- The task considered—predicting next step (forecasting) distribution, which is predominantly unimodal.\n \n I’d be curious to know whether the sampling methods remain applicable \n - for other tasks that might require modelling multimodal distributions: e.g. some sort of inverse problem (reconstruct field based on sparse, partial observations, compatible with multiple solutions), which is also a task of interest in dynamical system modelling, \n - or for other datasets: potentially still PDEs but not as coarsened and with more complicated nonlinear behaviour, or images. \n\n One could argue that Air_multi requires sampling from a more complicated distribution, and there the improvements are not as pronounced as in other settings. Perhaps you could include a brief discussion on this, or a preliminary experiment on a more complex PDE dataset, such as Kuramoto-Sivashinsky (see **Q1**)?\n2. **Counter-intuitive comparison to DDPM.** Maybe this is not necessarily a weakness, but I find it counter-intuitive that these sampling methods outperform DDPM, given that DDPM (or at least the continuous-time frame formulation) has stronger theoretical foundations (Similarly to how DDIM is posed as a framework where you can trade off computation for sample quality, not gain in both). I would have expected the same accuracy with significantly fewer NFEs to be possible, but not both. Is it possible that the DDPM model could be further tuned to achieve similar performance and maybe the DDPM baselines are not optimal? Maybe this is where a comparison with EDM would have been beneficial, as it provides a more principled approach of setting up diffusion models.\n3. **Lack of theoretical guarantees.** While the methods seem effective in practice, the paper does not provide any theoretical guarantees. I realise this might be hard to derive, but this makes the applicability of the methods more ad-hoc. This also reflects in the empirical investigation, where there is no clear recipe for what works best and how one should choose the hyperparameters optimally.\nAlso when making statements such as “However, we assert that\nour methods will consistently improve over ancestral sampling, as demonstrated in our experiments.” - This was only shown in three experiments, and the paper does not contain guarantees that this would always hold, so I would avoid over-generalising.\n4. **Lack of analysis of the stability of longer rollouts.** The paper provides several metrics to analyse the performance of the sampling methods, but none provides intuition/results about how the metrics evolve in time for the transient datasets (Tra and Fturb) (e.g. per-time-step MSE or per-time-step correlation). In particular, I’d be interested to compare the performance of the benchmarks to your methods on rollouts longer than what the model has been trained on. (This could, for instance, be tested on Tra_long from Kohl et al. [1]).\n \n I think that one potential weakness of these sampling schemes is that they lose more of the high frequency information than the traditional sampling schemes. Maybe this doesn’t affect the short rollouts significantly but it might negatively impact the accuracy of longer term rollouts (see Lippe et al. [3]). It would also be good to check that if extended significantly beyond the training range, the proposed sampling methods still generate physically plausible states, and outperform the baselines.\n5. **Lack of a comparison to EDM.** While I agree that a comparison to, for example, distillation techniques is outside the scope of the paper, I think EDM is relevant as a baseline. The fact that it is “designed to handle more complex stochastic data with multimodal distribution” does not mean it is not relevant for cases that do not exhibit much stochasticity. And it shouldn’t incur a different computational cost at training time. I think a comparison to EDM would be very useful to the community to figure out what the fastest and most accurate way to set up diffusion models for dynamical systems is. If the methods proposed here outperform EDM, this makes the paper stronger. If they don’t, I think this would also be a valuable result, potentially implying that EDM is a robust technique (regardless of data distribution) that should be used as a “first thing to try” as opposed to spending time and resources on hyperparameter tuning of more ad-hoc techniques.\n Maybe the authors could include a comparison to EDM on one or two experiments?\n6. **Unclear experimental setup.** I found it hard to figure out some experimental details in certain places. I am aware these details exist in other papers (Kohl et al. [1]), but for ease of interpreting the results, it would help to include some more details, potentially as a brief table in the appendix which summarises the most important dataset characteristics. For example: \n - Tra - What is the Mach number range of the training trajectories? Is it also Ma $\\in [0.53, 0.63] \\cup [0.69, 0.90]$ as in Kohl et al.?\n - Tra - How many trajectories are there per Mach number?\n - Fturb - I am slightly confused about the number of states within each trajectory for this dataset. You mention that each simulation contains 51 temporal states, but do you just consider 30 out of these for the test results? And what do you mean by “AR sampling is employed for $R = 30$ timesteps … for two sequences per $Re$?”\n - Fturb - you are feeding in the $Re$ number as conditioning information to the model as in Kohl et al., right?\n\n**Minor**\n\n7. **Comparison to DDIM (L306)** - You say that “IR should supersede the deterministic DDIM sampling regarding accuracy and NFEs” due to its stochastic nature. While I agree that stochastic sampling is beneficial because it can correct previous errors in sampling, Karras et al. [4] mention that, in practice, the situation is more complex because approximating the extra Langevin term introduces error in itself (see Section 4 Stochastic sampling). Thus, I would not say there is any guarantee that IR would supersede DDIM in all scenarios.\n8. **Lack of discussion on $j$.** I find it interesting that the optimum $j$ value varied so much between methods, as shown in Tables 6 and 7, but you do not include any discussion about this. Do you have any insight about why this might be the case?\n9. **Small typos**, such as L291 “to be evaluated”, L776 “presnted”, L772 “where” missing at the beginning of the line, L819 “allow” rather than “allows”, L840 “are not needed” etc.\n\nOverall, I think the paper is clearly written and structured, and generally presents convincing empirical evidence. However, I think its quality could be significantly improved by including experiments on longer rollouts, a comparison to EDM, and potentially clarifying the regimes in which the techniques are effective (with the inclusion of negative results if necessary).\n\n[4] Karras, T., Aittala, M., Aila, T., & Laine, S. (2022). Elucidating the Design Space of Diffusion-Based Generative Models. ArXiv, abs/2206.00364." }, "withdrawal_confirmation": null }, { "TLDR": { "value": "We propose an early truncation algorithm and an iterative refinement method to let diffusion models produce highly accurate results with low NFEs." }, "_bibtex": { "value": "@inproceedings{\nanonymous2024truncation,\ntitle={Truncation Is All You Need: Improved Sampling Of Diffusion Models For Physics-Based Simulations},\nauthor={Anonymous},\nbooktitle={Submitted to The Thirteenth International Conference on Learning Representations},\nyear={2024},\nurl={https://openreview.net/forum?id=0FbzC7B9xI},\nnote={under review}\n}" }, "abstract": { "value": "State-of-the-art Denoising Diffusion Probabilistic Models (DDPMs) rely on an expensive sampling process with a large Number of Function Evaluations (NFEs) to provide high-fidelity predictions. This computational bottleneck renders diffusion models less appealing as surrogates for the spatio-temporal prediction of physics-based problems with long rollout horizons. We propose Truncated Sampling Models, enabling single-step and few-step sampling with elevated fidelity by simple truncation of the diffusion process, reducing the gap between DDPMs and deterministic single-step approaches. We also introduce a novel approach, Iterative Refinement, to sample pre-trained DDPMs by reformulating the generative process as a refinement process with few sampling steps. Both proposed methods enable significant improvements in accuracy compared to DDPMs with NFEs $\\leq$ 10 on a diverse set of experiments, including incompressible and compressible turbulent flow and airfoil flow uncertainty simulations. Our proposed methods provide stable predictions for long rollout horizons in time-dependent problems and are able to learn all modes of the data distribution in steady-state problems with high uncertainty." }, "anonymous_url": { "value": "I certify that there is no URL (e.g., github page) that could be used to find authors’ identity." }, "authorids": null, "authors": null, "code_of_conduct": null, "code_of_ethics": { "value": "I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics." }, "comment": null, "confidence": null, "contribution": null, "desk_reject_comments": null, "details_of_ethics_concerns": null, "flag_for_ethics_review": null, "keywords": { "value": [ "physics-based simulations", "diffusion models", "improved sampling" ] }, "large_language_models": null, "no_acknowledgement_section": { "value": "I certify that there is no acknowledgement section in this submission for double blind review." }, "other_comments_on_LLMs": null, "paperhash": null, "pdf": { "value": "/pdf/487ead27e5e1a14dabe2352bcd3f95c789c0c375.pdf" }, "presentation": null, "primary_area": { "value": "applications to physical sciences (physics, chemistry, biology, etc.)" }, "questions": null, "rating": null, "reciprocal_reviewing": { "value": "I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6." }, "resubmission": null, "revert_desk_rejection_confirmation": null, "revert_withdrawal_confirmation": null, "soundness": null, "strengths": null, "student_author": null, "submission_guidelines": { "value": "I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide." }, "summary": null, "supplementary_material": { "value": "/attachment/ede03c067ebcd603195a870e18e6b1be7856a020.zip" }, "title": { "value": "Truncation Is All You Need: Improved Sampling Of Diffusion Models For Physics-Based Simulations" }, "venue": { "value": "ICLR 2025 Conference Submission" }, "venueid": { "value": "ICLR.cc/2025/Conference/Submission" }, "weaknesses": null, "withdrawal_confirmation": null } ]