allenai/tulu-3-IF-augmented-on-policy-70b
收藏Hugging Face2024-11-21 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/allenai/tulu-3-IF-augmented-on-policy-70b
下载链接
链接失效反馈官方服务：
资源简介：
Llama 3.1 Tulu 3 IF-Augmented数据集是一个偏好数据集，包含了来自SFT Data的提示，并结合了IFEval的约束，生成了65,530对生成对。这些生成对是通过多个模型生成的，包括Mistral、Tulu、Yi、MPT、Google Gemma、InternLM、Falcon、Qwen、Llama、GPT-4和Claude等。生成过程结合了on-policy和off-policy数据，并使用Ultrafeedback模板和LLM judge进行偏好标注。数据集主要用于研究和教育用途，遵循ODC-BY许可证。
Part of the Tulu 3 preference mixture dataset, containing 65,530 generation pairs generated by various models, combining prompts extracted from the SFT dataset and constraints from google/IFEval. The generation process combines on-policy and off-policy data, and preference annotations are made on four different aspects using the Ultrafeedback template and an LLM judge. The code for the dataset can be found in the scripts/synth_pref directory of the open-instruct project. The dataset is licensed under ODC-BY, intended for research and educational use, and follows Ai2s Responsible Use Guidelines.
提供机构：
allenai
