LLM4Hypergraph
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/iMoonLab/LLM4Hypergraph
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为LLM4Hypergraph，是一个包含21,500个问题的综合性基准测试，涵盖了八个低阶、五个高阶以及两个同构任务，同时使用了来自引文网络和蛋白质结构的合成与真实超图。该基准测试能够评估模型在理解复杂超图关系方面的优势和不足。任务旨在对超图理解任务上的LLM模型进行评估。
The dataset, named LLM4Hypergraph, is a comprehensive benchmark containing 21,500 questions. It encompasses 8 low-order, 5 high-order, and 2 isomorphism tasks, utilizing both synthetic and real hypergraphs sourced from citation networks and protein structures. This benchmark can evaluate the strengths and limitations of models in comprehending complex hypergraph relationships, with its tasks specifically designed to assess Large Language Models (LLMs) on hypergraph understanding tasks.
提供机构：
iMoonLab
