BubbleRAG: Evidence-Driven Retrieval-Augmented Generation for Black-Box Knowledge Graphs

TL;DR

BubbleRAG is a novel, training-free retrieval method that enhances evidence discovery in black-box knowledge graphs, significantly improving accuracy and F1 scores in multi-hop question answering tasks.

Contribution

It formalizes the retrieval problem as OISR, introduces BubbleRAG with innovative techniques, and demonstrates state-of-the-art performance without additional training.

Findings

Outperforms strong baselines in F1 and accuracy

Effectively balances recall and precision in evidence retrieval

Operates as a plug-and-play solution for knowledge graph tasks

Abstract

Large Language Models (LLMs) exhibit hallucinations in knowledge-intensive tasks. Graph-based retrieval augmented generation (RAG) has emerged as a promising solution, yet existing approaches suffer from fundamental recall and precision limitations when operating over black-box knowledge graphs -- graphs whose schema and structure are unknown in advance. We identify three core challenges that cause recall loss (semantic instantiation uncertainty and structural path uncertainty) and precision loss (evidential comparison uncertainty). To address these challenges, we formalize the retrieval task as the Optimal Informative Subgraph Retrieval (OISR) problem -- a variant of Group Steiner Tree -- and prove it to be NP-hard and APX-hard. We propose BubbleRAG, a training-free pipeline that systematically optimizes for both recall and precision through semantic anchor grouping, heuristic bubble expansion to discover candidate evidence graphs (CEGs), composite ranking, and reasoning-aware expansion. Experiments on multi-hop QA benchmarks demonstrate that BubbleRAG achieves state-of-the-art results, outperforming strong baselines in both F1 and accuracy while remaining plug-and-play.