Aurora-M: The First Open Source Multilingual Language Model Red-teamed according to the U.S. Executive Order
Paper
•
2404.00399
•
Published
•
41
Aurora-M models (base, biden-harris redteams and instruct)
Note Base model trained on Japanese, Vietnamese, Hindi, English and Finnish and some instruction data mixed into pretraining.
Note A finetuned version of the base model on a bunch of instruction tuning datasets and the biden-harris redteam dataset.
Note A finetuned version of the base model just trained on slim-orca-dedup.
Note Biden Harris redteam dataset