NinaCalvi/ultra-rm-truthfulness-1000-Skywork

该数据集包含多个字段，涵盖了指令、完成情况、评分、理由、模型、系统提示等内容。具体字段包括source、instruction、completions、correct_answers、incorrect_answers、split、annotations等。completions字段下的annotations包含helpfulness、honesty、instruction_following、truthfulness等评分和理由。数据集还包含训练集的分割信息，如train分割的字节数和示例数。

This dataset contains multiple fields covering instructions, completions, ratings, rationales, models, system prompts, and more. Specific fields include source, instruction, completions, correct_answers, incorrect_answers, split, annotations, etc. The annotations field under completions includes ratings and rationales for helpfulness, honesty, instruction_following, and truthfulness. The dataset also includes split information for the training set, such as the number of bytes and examples in the train split.