allenai/tulu-3-sft-reused-off-policy

该数据集是Tulu 3偏好混合的一部分，包含来自SFT混合的提示和96,911个生成对，这些生成对使用了多个模型，包括Mistral、Tulu、Yi、MPT、Google Gemma、InternLM、Falcon、Qwen、Llama、GPT-4和Claude等。数据集的特征包括id、prompt、chosen和rejected，其中chosen和rejected是包含content和role的列表。数据集的分割仅包含训练集，大小为584,556,391字节，包含96,911个示例。数据集遵循ODC-BY许可，主要用于研究和教育用途，并遵循Ai2的负责任使用指南。

This is a preference dataset, part of the Tulu 3 preference mixture. It contains prompts extracted from the SFT mixture and includes 96,911 generation pairs produced using various models. The dataset features include id, prompt, chosen, and rejected, where chosen and rejected each contain content and role. The training portion of the dataset contains 96,911 samples with a total size of 584,556,391 bytes. The download size of the dataset is 301,118,275 bytes. The dataset is licensed under ODC-BY, intended for research and educational use, and follows Ai2s Responsible Use Guidelines. The dataset includes output data from third-party models, which are subject to their respective terms of use.