Evidence-guided retrieval improves multi-hop question answering accuracy
EAGR: Evidence-Guided Adaptive Graph Retrieval
Information Retrieval
Summary
Multi-hop question answering means answering questions that need multiple pieces of information linked together. The authors introduce a method called EAGR where the system dynamically decides which pieces of information to look at next based on what it has already found. This approach does not need complicated graph databases and avoids repeatedly changing the search query. The authors show that EAGR improves accuracy on three popular multi-step question answering tests, beating existing methods by a notable margin.
What this means in practice
- •For natural language processing teams: Enhance multi-step question answering systems by dynamically guiding retrieval based on intermediate answers without building explicit graph databases.
- •For enterprise search engineers: Improve document retrieval in complex search tasks by using evidence-conditioned semantic exploration within standard vector databases.
Authors
Sonam Gupta, Vignesh P, Vishwajeet Kumar, Adnan Qidwai, Anand Eswaran, Jaydeep Sen, Sachindra Joshi
Abstract
Retrieval-Augmented Generation (RAG) for multi-hop reasoning requires retrieval to adapt as intermediate evidence reveals new information needs. Existing approaches support such retrieval through explicit graph structures or iterative query reformulation, introducing graph-construction or query-time overhead. We present Evidence-Guided Adaptive Graph Retrieval (EAGR), a framework in which accumulated evidence dynamically determines which semantic directions to explore next, and which requires no dedicated graph database and no explicit graph materialization. EAGR operates over an implicit semantic graph induced by compact passage metadata pre-ingested into a standard VectorDB. Starting from dense-retrieval seed passages, an LLM iteratively assesses accumulated evidence, identifies promising semantic directions, expands the retrieval frontier through metadata-based neighborhood lookup, and terminates once sufficient evidence has been collected. By precomputing semantic metadata, EAGR enables evidence-conditioned graph exploration without materializing an explicit graph, maintaining a dedicated GraphDB, or iteratively rewriting the query. Across three widely used multi-hop QA benchmarks, EAGR achieves the highest LLM-based answer accuracy among evaluated methods, outperforming the vanilla RAG by 17\% and strongest graph-aware baseline by 4\% percentage points averaged across three popular benchmarks HotPotQA, 2WikiMultiHopQA, and MuSiQue.