Query-Aware Multi-Relational Evidence Graph for Retrieval-Augmented Generation

Abstract

Retrieval-Augmented Generation (RAG) systems typically apply identical retrieval logic to all user queries, overlooking that different query intents depend on fundamentally different document relationships. We propose QMEG, a query-aware retrieval framework that constructs a multi-relational evidence graph with five active semantic edge types (plus one reserved for future borrowing data) and dynamically activates edge subsets according to query intent. On a library dataset of 19,818 books with 250 annotated queries, QMEG achieves 31% higher diversity on learningpath queries and 47% higher LLM-judged quality on fact queries compared to fixed-graph baselines, while remaining robust under weight perturbation. Ablation reveals that the topic-similarity edge is indispensable for exploratory recommendation, with diversity dropping 34% upon its removal.