swiss-ai/realtoxicityprompts

这是一个包含两个配置（realtoxicityprompts_full和realtoxicityprompts_small）的数据集，用于研究文本中的毒性。每个配置都包括训练集，数据集中的每个样本都包含了文本的开始和结束位置、是否具有挑战性的标记、文本内容以及不同类型毒性（如调情、身份攻击、侮辱、咒骂、严重毒性、露骨的性内容、威胁和毒性）的评分。

This is a dataset with two configurations (realtoxicityprompts_full and realtoxicityprompts_small) designed for studying toxicity in text. Each configuration includes a training set, and each sample in the dataset contains the start and end positions of the text, a challenging flag, the text content, and ratings for different types of toxicity such as flirtation, identity attack, insult, profanity, severe toxicity, sexually explicit content, threat, and toxicity.