Insight at the right spot: Provide decisive subgraph information to Graph LLM with reinforcement learning

Tiesunlong Shen, Erik Cambria, Jin Wang, Yi Cai, Xuejie Zhang

Abstract: Highlights•A novel RL-based approach enhancing Graph LLMs with decisive subgraph information.•Adaptive RL module identifies essential neighborhoods for LLM predictions on graphs.•Experiments on five datasets show Spider outperforms existing Graph LLM methods.