Xiao-Yu Guo

Graph Evidence Is Not Enough: Diagnosing Native Decoder Use in Graph-Augmented LLMs

This work presents S$^2$GE as an instance showing that diagnosis-driven interface design can improve native decoder usability and introduces an intervention triangle with three matched conditions: readable graph evidence, shuffled graph evidence, and no-graph input that separates evidence inclusion, structural readability, and decoder-usable topology.