Hierarchical Lexical Graph for Enhanced Multi-Hop Retrieval

TL;DR

This paper introduces a hierarchical lexical graph (HLG) to improve multi-hop retrieval in retrieval-augmented generation, enabling better connection of distant documents and more accurate answers.

Contribution

It proposes a novel three-tier index structure and two plug-and-play retrievers, along with a synthetic dataset for robust evaluation of multi-hop retrieval systems.

Findings

Outperforms naive chunk-based RAG with 23.1% improvement in recall and correctness

Introduces a synthetic dataset for multi-hop retrieval evaluation

Demonstrates effectiveness across five datasets

Abstract

Retrieval-Augmented Generation (RAG) grounds large language models in external evidence, yet it still falters when answers must be pieced together across semantically distant documents. We close this gap with the Hierarchical Lexical Graph (HLG), a three-tier index that (i) traces every atomic proposition to its source, (ii) clusters propositions into latent topics, and (iii) links entities and relations to expose cross-document paths. On top of HLG we build two complementary, plug-and-play retrievers: StatementGraphRAG, which performs fine-grained entity-aware beam search over propositions for high-precision factoid questions, and TopicGraphRAG, which selects coarse topics before expanding along entity links to supply broad yet relevant context for exploratory queries. Additionally, existing benchmarks lack the complexity required to rigorously evaluate multi-hop summarization systems, often focusing on single-document queries or limited datasets. To address this, we introduce a synthetic dataset generation pipeline that curates realistic, multi-document question-answer pairs, enabling robust evaluation of multi-hop retrieval systems. Extensive experiments across five datasets demonstrate that our methods outperform naive chunk-based RAG achieving an average relative improvement of 23.1% in retrieval recall and correctness. Open-source Python library is available at https://github.com/awslabs/graphrag-toolkit.