ustc-zhangzm/BoolQuestions
收藏Hugging Face2024-12-09 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/ustc-zhangzm/BoolQuestions
下载链接
链接失效反馈官方服务：
资源简介：
BoolQuestions数据集旨在研究密集检索模型是否理解语言中的布尔逻辑。该数据集包含来自NaturalQuestions和MSMARCO的数据，分为eval和corpus两种配置。eval配置包含问题、问题类型、正面上下文和负面上下文等特征，而corpus配置包含文档ID和文档内容等特征。数据集可以通过Hugging Face Hub或datasets库加载，并提供了评估代码的GitHub链接。
BoolQuestions is a dataset designed to evaluate whether dense retrieval models understand Boolean logic in language. It includes source data from NaturalQuestions and MSMARCO, divided into evaluation sets and corpora. The dataset features include questions, question types, positive and negative contexts, etc. The purpose of the dataset is for research purposes and may contain some noise.
提供机构：
ustc-zhangzm
