AlignmentResearch/Harmless
收藏Hugging Face2024-07-29 更新2024-07-13 收录
下载链接：
https://hf-mirror.com/datasets/AlignmentResearch/Harmless
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含多个配置（default、neg、pos），每个配置都有相同的特征结构，包括instructions（指令）、content（内容）、answer_prompt（答案提示）、clf_label（分类标签）和gen_target（生成目标）。分类标签clf_label有两个类别：FIRST和SECOND。数据集分为训练集和验证集，每个配置的训练集和验证集的大小和样本数量有所不同。
The dataset contains multiple configurations (default, neg, pos), each with the same feature structure, including instructions, content, answer_prompt, clf_label, and gen_target. The classification label clf_label has two categories: FIRST and SECOND. The dataset is divided into training and validation sets, with varying sizes and numbers of samples for each configuration.
提供机构：
AlignmentResearch
