risky-conversations/jailbreaks_dataset_with_results_reduced

这是一个包含会话信息的数据集，每个会话包含文本内容、角色信息以及多个维度的毒性评分，如身份攻击、侮辱、粗俗等。同时，提供了会话的长度、连贯性、可读性等指标，以及数据集的训练集划分和相关的配置信息。

This dataset contains conversation information, with each conversation including text content, role information, and toxicity scores across multiple dimensions such as identity attack, insult, obscenity, etc. It also provides metrics such as conversation length, coherence, readability, and cognitive load, as well as the datasets training set split and related configuration information.