SaketR1/civil_comments_preference_pairs
收藏官方服务:
资源简介:
这个数据集包含了用于检测文本毒性的样本,每个样本由一个提示和两个可能的完成句子组成,一个是正面的,另一个是负面的。毒性分数是一个从0到1的浮点数,表示文本的毒性程度。数据集提供了一个训练集用于模型训练。
The dataset consists of samples for detecting text toxicity, each with a prompt and two possible completion sentences, one favorable and one unfavorable. The toxicity score is a floating-point number between 0 and 1, indicating the level of toxicity of the text. The dataset provides a training set for model training.
提供机构:
SaketR1


