Bridging the Completeness Gap in GraphRAG: A Unified GNN Enhancement Layer for Retrieval and Generation

GraphRAG retrieves evidence from knowledge graphs automatically constructed from text, yet dense graphs can still omit the relations that downstream queries need: on GraphRAG-Bench Medical, a graph with an average degree of $6.86$ covers only $14.5\%$ of annotated evidence triples, and in $61.0\%$ of cases the relevant entities are present but the relation between them is missing. We introduce a lightweight query-adaptive GNN layer that predicts missing relations, estimates their confidence and query relevance, and selects them for structural augmentation and verbal hints. A structural path inserts query-adaptively filtered relations into the graph before Personalized PageRank (PPR) retrieval, while a verbal path passes top-ranked relations to the language model as concise hints, exploiting relational knowledge without propagating noisy edges through the graph. Filtering is decisive: inserting all predicted edges lowers Medical accuracy by $5.23$ points, whereas a learned query-adaptive gate turns this into a $3.47$-point gain. The verbal path adds $4.95$ points, and the two paths together gain $7.94$ and $5.54$ points on Medical and Novel, respectively. The layer improves six retrieval-system configurations across the evaluated benchmark subsets and three language-model backbones on Medical, supporting selective structural and verbal augmentation as a practical way to address missing connections in this setting.