Graph-Anchored Knowledge Indexing for Retrieval-Augmented Generation

2026-01-23

AI Summary

This work addresses the challenge faced by existing retrieval-augmented generation (RAG) systems in effectively integrating critical evidence scattered across noisy documents for multi-hop question answering. To this end, the authors propose GraphAnchor, a novel approach that leverages a graph structure as a dynamically evolving knowledge index. During iterative retrieval, GraphAnchor incrementally constructs and anchors key entities and their relationships, guiding large language models to assess knowledge sufficiency, generate sub-queries, and jointly optimize retrieval and generation. By employing graph-driven attention modulation, the method provides structured guidance along multi-hop reasoning paths. Experimental results demonstrate that GraphAnchor significantly outperforms current state-of-the-art methods on four multi-hop QA benchmarks, confirming its capability to dynamically focus on dispersed yet essential information.

Technology Category

• Natural Language Processing: Question Answering
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Machine Learning: Graph-based Machine Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Retrieval-Augmented Generation (RAG) has emerged as a dominant paradigm for mitigating hallucinations in Large Language Models (LLMs) by incorporating external knowledge. Nevertheless, effectively integrating and interpreting key evidence scattered across noisy documents remains a critical challenge for existing RAG systems. In this paper, we propose GraphAnchor, a novel Graph-Anchored Knowledge Indexing approach that reconceptualizes graph structures from static knowledge representations into active, evolving knowledge indices. GraphAnchor incrementally updates a graph during iterative retrieval to anchor salient entities and relations, yielding a structured index that guides the LLM in evaluating knowledge sufficiency and formulating subsequent subqueries. The final answer is generated by jointly leveraging all retrieved documents and the final evolved graph. Experiments on four multi-hop question answering benchmarks demonstrate the effectiveness of GraphAnchor, and reveal that GraphAnchor modulates the LLM's attention to more effectively associate key information distributed in retrieved documents. All code and data are available at https://github.com/NEUIR/GraphAnchor.

Problem

• Retrieval-Augmented Generation
• hallucination mitigation
• knowledge integration
• noisy documents
• multi-hop question answering

Innovation

• Graph-Anchored Knowledge Indexing
• Retrieval-Augmented Generation
• Dynamic Knowledge Graph
• Iterative Retrieval
• Multi-hop Question Answering