Update README.md

README.md

@@ -7,7 +7,7 @@ tags:
 - llama
 - llama-2
 ---
-[WIP]
+[Probeversion]
 
 This is the LLaMAfied version of [Qwen/Qwen-7B-Chat](https://huggingface.co/Qwen/Qwen-7B-Chat), recalibrated to fit the original LLaMA/LLaMA-2-like model structure.
 
@@ -15,15 +15,19 @@ You can use LlamaForCausalLM for model inference, which is the same as LLaMA/LLa
 
 The model has been edited to be white-labelled, meaning the model will no longer call itself a Qwen.
 
-SPOILOR: Further finetuning is in progress, the current version is a work-in-progress, some knowledge may be biased and illusory due to structural changes. Will be updated very, very sooooooooooon.
+SPOILOR: Further finetuning is in progress, the current version is a work-in-progress, some knowledge may be biased and illusory due to structural changes. Will be updated very, very soooooooooon.
 
 PROMPT FORMAT: [chatml](https://github.com/openai/openai-python/blob/main/chatml.md)
 
-CURRENT MMLU: 50.36
+CURRENT MMLU: 53.48
 
-Issue: Compared to the original Qwen-Chat scoring 53.9, the MMLU score dropped slightly (-3.54) due to insufficient realignment.
+```
+stem ACC: 46.40 Humanities ACC: 47.61 other ACC: 61.31 social ACC: 61.78 AVERAGE ACC:53.48
+```
 
-[在制品]
+Issue: Compared to the original Qwen-7B-Chat scoring 53.90, the MMLU score dropped slightly (-0.42) due to insufficient realignment.
+
+[预览版本]
 
 这是 [通义千问 Qwen/Qwen-7B-Chat](https://huggingface.co/Qwen/Qwen-7B-Chat) 的 LLaMA 化版本，经过重新校准以适应原始的类似 LLaMA/LLaMA-2 的模型结构。
 
@@ -31,10 +35,14 @@ Issue: Compared to the original Qwen-Chat scoring 53.9, the MMLU score dropped s
 
 模型已经被编辑实现白标化，不再自称通义千问。
 
-剧透： 进一步的微调正在进行中，当前版本是一个正在进行的工作，一些知识可能由于结构变化而产生偏见和幻觉。 会更新，很快，非常非常非常快。
+剧透： 进一步的微调正在进行中，当前版本是一个正在进行的工作，一些知识可能由于结构变化而产生偏见和幻觉。 会更新，很快，非常非常快。
 
 PROMPT 格式: [chatml](https://github.com/openai/openai-python/blob/main/chatml.md)
 
-当前的 MMLU: 50.36
+当前的 MMLU: 53.48
+
+```
+stem ACC: 46.40 Humanities ACC: 47.61 other ACC: 61.31 social ACC: 61.78 AVERAGE ACC:53.48
+```
 
-问题：相比原本的Qwen-Chat的53.9，由于不够充分的重新对齐，MMLU分数略有下降（-3.54）。
+问题：相比原本的 Qwen-7B-Chat 的 53.90，由于不够充分的重新对齐，MMLU分数略有下降（-0.42）。