OntoURL is a benchmark for evaluating large language models on symbolic ontological understanding, reasoning, and learning. The benchmark is generated from formal ontology resources and covers closed-form question answering, reasoning over ontology-derived knowledge, and structured ontology construction tasks.
36,159 benchmark instances
15 tasks covering understanding, reasoning, and learning
43 source ontology resources across 8 domains
Multiple… See the full description on the dataset page:
https://huggingface.co/datasets/XiaoZhang98/OntoURL.
