nayohan/KoSBi

该数据集名为KoSBi，旨在通过缓解社会偏见风险来提高大型语言模型应用的安全性。数据集包含人口统计类别、人口统计组、上下文、英文上下文、句子、英文句子、上下文标签、上下文子标签、句子标签、句子子标签和注释的人口统计信息等特征。数据集分为训练集、验证集和测试集，分别包含54393、6799和6801个示例。数据集的许可证为MIT，主要语言为韩语，标签为安全。

This dataset aims to mitigate social bias risks in large language model applications by providing various features such as demographic category, demographic group, context, sentence, etc. The dataset is divided into train, valid, and test splits, each with specified number of examples and bytes. The download size and total size of the dataset are also detailed, along with configurations, license, and language information.