BSC-LT/EsBBQ

EsBBQ（西班牙语问题回答偏见基准）是一个用于评估语言模型中社会偏见的西班牙语数据集。它包括多个社会类别，如年龄、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。该数据集包含各种特征和分割，旨在通过多选题问题回答设置来评估模型在社会偏见方面的表现。它包括具有歧义和清晰信息的上下文，以评估模型是否依赖于社会偏见并忽略正确答案。

The Spanish Bias Benchmark for Question Answering (EsBBQ) is a dataset designed to evaluate social biases in language models. It includes various features and splits, organized into different social categories such as Age, Disability Status, Gender, LGBTQIA, Nationality, Physical Appearance, Race/Ethnicity, Religion, Socioeconomic Status (SES), and Spanish Region. The dataset is structured to assess model performance in a multiple-choice Question Answering (QA) setting, with contexts that are either ambiguous or adequately informative to evaluate the degree of reliance on social biases and the ability to disregard the correct answer.