allenai/tulu-3-wildchat-if-on-policy-8b

Llama 3.1 Tulu 3 Wildchat IF数据集是一个偏好数据集，包含来自WildChat的提示以及使用多种模型生成的10,792对生成对。这些生成对是通过合成管道生成的，结合了策略内和策略外的数据，并使用Ultrafeedback模板和LLM法官在四个不同方面进行了偏好注释。数据集的特征包括id、prompt、chosen和rejected，其中chosen和rejected分别包含content和role字段。数据集的分割为train，包含90,414,403字节和10,792个示例。数据集的大小为90,414,403字节，下载大小为46,225,965字节。数据集的许可证为ODC-BY，适用于研究和教育用途。

This preference dataset is part of our Tulu 3 preference mixture. It contains prompts from WildChat, which include constraints, and it contains 10,792 generation pairs (some of which on-policy from allenai/Llama-3.1-Tulu-3-8B) obtained using various models including Mistral 7B Instruct v0.2, Tulu 2 7B, Yi-34B-Chat, etc. The dataset generation process combines both on-policy and off-policy data, and uses the Ultrafeedback template and an LLM judge for preference annotations. The code for the synthetic generation pipeline is 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.