2 papers
cs.LG2026
LLM Features Can Hurt GNNs: Concatenation Interference on Homophilous Graph Benchmarks
Zhongyuan Wang, Pratyusha Vemuri
Adding LLM-generated node features to graph neural networks (GNNs) is widely reported to improve accuracy on standard benchmarks. We document a contrasting observation: when LLM fe…
cs.AI2026
When the Tool Decides: LLM Agents Defer Blindly to Graph Neural Network Tools, and Stronger Backbones Defer More
Zhongyuan Wang, Pratyusha Vemuri
A growing line of work equips large language model (LLM) agents with graph neural networks (GNNs) as callable tools, assuming the agent exercises judgment over when and how much to…