Knowledge Graph-Enhanced Large Language Models via Path Selection

Two datasets, MetaQA and FACTKG, are used to evaluate the effectiveness of the proposed method KELP. MetaQA is a critical benchmark dataset containing subsets of questions with 1-hop/2-hop reasoning steps, and featuring a wide variety of questions. FACTKG is a dataset comprising 108,000 claims categorized as either True or False, where claims are subject to validation with DBpedia.