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

Llama 3.1 Tulu 3 Wildchat数据集是一个偏好数据集，包含来自WildChat的提示和17,207个生成对。这些生成对使用了多个不同的模型生成，包括Mistral 7B Instruct v0.2、Tulu 2 7B、Yi-34B-Chat等。数据集的特征包括id、prompt、chosen和rejected，其中chosen和rejected是包含内容和角色的列表。数据集的分割仅包含训练集，大小为141,797,370字节，包含17,207个示例。数据集的生成方法结合了on-policy和off-policy数据，并使用Ultrafeedback模板和LLM法官进行偏好注释。数据集遵循ODC-BY许可证，适用于研究和教育用途。

This preference dataset is part of the Tulu 3 preference mixture, containing prompts from WildChat and 17,207 generation pairs. The generations were produced using various models including Mistral 7B Instruct v0.2, Tulu 2 7B, Yi-34B-Chat, etc. The dataset was generated using a synthetic pipeline combining on-policy and off-policy data, with preference annotations made using the Ultrafeedback template and an LLM judge. The dataset is licensed under ODC-BY and is intended for research and educational use.