aladinDJ/ultrafeedback-DPO-annotated
收藏Hugging Face2025-11-14 更新2025-11-15 收录
下载链接:
https://hf-mirror.com/datasets/aladinDJ/ultrafeedback-DPO-annotated
下载链接
链接失效反馈官方服务:
资源简介:
这个数据集提供了原始Ultrafeedback偏好语料库的Magpie注释元数据,包括任务类别、难度、输入质量、语言和独立的偏好奖励分数。它被用作UltraMix精选管道中五个基础组件之一。
This dataset provides Magpie-annotated metadata for the original Ultrafeedback preference corpus, including task category, difficulty, input quality, language, and independent preference reward scores. It serves as one of the five foundational components used in the UltraMix curation pipeline.
提供机构:
aladinDJ



