BSC-LT/CaBBQ

CaBBQ是一个用于评估语言模型中社会偏见的西班牙语和加泰罗尼亚语数据集。它基于原始的BBQ数据集，并针对西班牙的社会背景进行了调整。该数据集在多个选择问答（QA）环境中对10个社会类别进行评估，包括年龄、残疾状况、性别、LGBTQIA、国籍、外貌、种族/民族、宗教、社会经济状况（SES）和西班牙地区。每个实例都包含关于模板、上下文、问题、答案和社会群体的详细信息。

CaBBQ is a Spanish and Catalan dataset for evaluating social biases in language models. It is an adaptation of the original BBQ dataset tailored to the social context of Spain. The dataset assesses biases across 10 social categories in a multiple-choice Question Answering (QA) setting, including Age, Disability Status, Gender, LGBTQIA, Nationality, Physical Appearance, Race/Ethnicity, Religion, Socioeconomic Status (SES), and Spanish Region. Each instance includes detailed information about templates, contexts, questions, answers, and social groups.