Magpie-Align/Magpie-100K-Generator-Zoo
收藏Hugging Face2024-12-09 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/Magpie-Align/Magpie-100K-Generator-Zoo
下载链接
链接失效反馈官方服务：
资源简介：
该数据集用于论文《Stronger Models are NOT Stronger Teachers for Instruction Tuning》。数据集的指令从Magpie-Air中采样，并使用19种不同的响应生成器生成响应。数据集包含uuid、instruction以及多个模型生成的响应内容及其对应的奖励值。可以基于提供的奖励值构建DPO数据集。
This dataset is used by the paper Stronger Models are NOT Stronger Teachers for Instruction Tuning. Instructions are sampled from Magpie-Air, and responses are generated using 19 different response generators. The dataset includes uuid, instruction, and responses generated by multiple models along with their corresponding reward values. A DPO dataset can be built based on the provided reward values.
提供机构：
Magpie-Align
