EvoPath: Evolutionary meta-path discovery with large language models for complex heterogeneous information networks

Shixuan Liu, Haoxiang Cheng, Yunfei Wang, Yue He, Changjun Fan, Zhong Liu

Abstract: Highlights•We use LLMs to discover meta-paths for better reasoning in complex HINs.•Our framework resolves conflicts between LLMs’ parametric knowledge and HINs’ contextual knowledge.•Experiments show that our generated meta-paths excel in reasoning tasks.