FastInsight: Fast and Insightful Retrieval via Fusion Operators for Graph RAG

TL;DR

FastInsight introduces a novel, efficient graph retrieval method that combines graph and vector search techniques to improve retrieval accuracy and generation quality in large language model applications.

Contribution

It proposes a new taxonomy of graph retrieval operations and introduces two fusion operators, GRanker and STeX, to address limitations in existing methods.

Findings

Significantly improves retrieval accuracy over baselines

Enhances generation quality in large language model tasks

Achieves better effectiveness-efficiency trade-offs

Abstract

Existing Graph RAG methods aiming for insightful retrieval on corpus graphs typically rely on time-intensive processes that interleave Large Language Model (LLM) reasoning. To enable time-efficient insightful retrieval, we propose FastInsight. We first introduce a graph retrieval taxonomy that categorizes existing methods into three fundamental operations: vector search, graph search, and model-based search. Through this taxonomy, we identify two critical limitations in current approaches: the topology-blindness of model-based search and the semantics-blindness of graph search. FastInsight overcomes these limitations by interleaving two novel fusion operators: the Graph-based Reranker (GRanker), which functions as a graph model-based search, and Semantic-Topological eXpansion (STeX), which operates as a vector-graph search. Extensive experiments on broad retrieval and generation datasets demonstrate that FastInsight significantly improves both retrieval accuracy and generation quality compared to state-of-the-art baselines, achieving a substantial Pareto improvement in the trade-off between effectiveness and efficiency.