MINTQA
收藏资源简介:
MINTQA是由爱丁堡大学、东南大学和曼彻斯特大学联合创建的多跳问答基准数据集,旨在评估大型语言模型在处理复杂、知识密集型多跳查询中的能力。数据集包含10,479个新知识问题和17,887个长尾知识问题,涵盖了从1跳到4跳的复杂推理任务。数据集通过从Wikidata和Wikipedia中提取知识三元组,并使用GPT-4o生成多跳问题,确保了问题的多样性和复杂性。MINTQA的应用领域主要集中在多跳问答任务中,旨在解决模型在处理新知识、长尾知识以及复杂推理任务时的不足。
MINTQA is a multi-hop question answering benchmark dataset jointly created by the University of Edinburgh, Southeast University, and the University of Manchester, aiming to evaluate the capabilities of large language models when handling complex, knowledge-intensive multi-hop queries. It consists of 10,479 newly emerged knowledge questions and 17,887 long-tail knowledge questions, covering complex reasoning tasks ranging from 1-hop to 4-hop. The dataset is constructed by extracting knowledge triples from Wikidata and Wikipedia, and generating multi-hop questions using GPT-4o, which ensures the diversity and complexity of the questions. The main application scenario of MINTQA focuses on multi-hop question answering tasks, and it is designed to address the shortcomings of models when dealing with newly emerged knowledge, long-tail knowledge, and complex reasoning tasks.




