sagepond/instructions_kto_v2
收藏Hugging Face2025-08-18 更新2025-11-02 收录
下载链接：
https://hf-mirror.com/datasets/sagepond/instructions_kto_v2
下载链接
链接失效反馈官方服务：
资源简介：
Instructions KTO是一个专门为训练和评估人类感知损失函数而设计的数据集，包含人类指令-响应对及二元反馈标签，共有大约28k个示例，分为训练集和测试集。
Instructions KTO is a dataset designed for training and evaluating human-aware loss functions, containing human instruction-response pairs with binary feedback labels, totaling about 28k examples, split into training and test sets.
提供机构：
sagepond
