Large Language Models Still Face Challenges in Multi-Hop Reasoning with External Knowledge

Haotong Zhang

We carry out a series of experiments to test large language models' multi-hop reasoning ability from three aspects: selecting and combining external knowledge, dealing with non-seq…