WorldTree

WorldTree是一个专为小学科学问题设计的数据集，包含约1680个标准化科学考试问题的详细解释，以解释图的形式呈现。数据集由亚利桑那大学创建，旨在通过提供多跳推理支持，帮助训练可解释的推理模型。数据集内容包括解释图和解释中心表格存储，后者包含62个半结构化表格，共4950行，用于构建科学解释。该数据集主要应用于科学教育领域，解决自动化推理系统中解释生成的问题，提高用户信任和错误检测能力。

WorldTree is a dataset specifically designed for elementary school science questions, containing detailed explanations for approximately 1,680 standardized science exam questions presented in the form of explanation graphs. Developed by the University of Arizona, this dataset aims to support multi-hop reasoning and facilitate the training of interpretable reasoning models. The dataset includes two core components: explanation graphs and explanation-centric table repositories, where the latter consists of 62 semi-structured tables with a total of 4,950 rows that are used for constructing scientific explanations. Primarily applied in the field of science education, this dataset addresses the problem of explanation generation in automated reasoning systems, and enhances user trust and error detection capabilities.