RiTeK
收藏资源简介:
RiTeK是一个专为大型语言模型在文本知识图谱上进行复杂推理而设计的数据集,主要应用于医疗领域。该数据集包含15557条数据,涵盖了丰富的拓扑结构和关系类型,旨在模拟真实世界的医疗查询场景。数据集的创建过程包括专家设计的拓扑结构模板、文本属性的提取与整合,以及专家对合成查询的严格评估。RiTeK的应用领域主要集中在医疗诊断和治疗计划的复杂查询解答,旨在提高大型语言模型在处理复杂医疗问题上的推理能力。
RiTeK is a dataset specifically designed for complex reasoning over textual knowledge graphs by large language models, primarily targeting the medical domain. It contains 15,557 data instances, which incorporate a wide range of topological structures and relationship types, and is intended to simulate real-world medical query scenarios. The dataset's development process includes expert-designed topological structure templates, extraction and integration of textual attributes, as well as rigorous expert evaluation of the synthesized queries. The main application areas of RiTeK are complex query answering for medical diagnosis and treatment planning, aiming to enhance the reasoning capabilities of large language models when handling complex medical issues.

- 1RiTeK: A Dataset for Large Language Models Complex Reasoning over Textual Knowledge Graphs马萨诸塞大学洛厄尔分校、马萨诸塞大学阿默斯特分校、明尼苏达大学、埃默里大学、马萨诸塞大学医学院 · 2024年



