GraphextQA

GraphextQA是由慕尼黑大学创建的一个问题回答数据集，包含59,964对问题和相应的子图，数据来源于Wikidata。该数据集旨在评估和促进图语言模型的发展，特别是模型理解和利用图信息生成答案的能力。数据集中的问题涉及广泛的真实世界知识，子图提供了从问题中提到的实体到答案实体的推理路径。GraphextQA不仅用于评估模型，还支持开发能够将图知识集成到语言模型中的算法。

GraphextQA is a question answering dataset developed by Ludwig-Maximilians-Universität München (LMU Munich). It contains 59,964 question-answer pairs, each paired with a corresponding subgraph, with all data sourced from Wikidata. This dataset is designed to evaluate and advance the development of graph language models, particularly the ability of models to understand and leverage graph information to generate answers. The questions in the dataset cover a wide range of real-world knowledge, while the subgraphs provide the reasoning paths from the entities mentioned in the questions to the answer entities. Beyond being used for model evaluation, GraphextQA also supports the development of algorithms that integrate graph knowledge into language models.