Towards Open-World Retrieval-Augmented Generation on Knowledge Graph: A Multi-Agent Collaboration Framework

2025-09-01

AI Summary

Existing KG-RAG methods rely on predefined anchor entities to initiate graph traversal, rendering them insufficiently robust for open-world question answering due to unreliable entity linking. This paper proposes the first anchor-free, multi-agent collaborative KG-RAG framework, comprising prediction, retrieval, and supervision agents. It enables end-to-end robust knowledge exploration and answer generation via semantic alignment–driven parallel multi-hop graph traversal, dynamic policy control, and iterative retrieval refinement—eliminating explicit entity linking entirely. Consequently, the method exhibits markedly improved adaptability to ambiguous or erroneous queries. Evaluated on four public benchmarks, it consistently outperforms state-of-the-art (SOTA) approaches; moreover, it establishes new SOTA performance on real-world question-answering tasks.

Technology Category

• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Multiagent Systems: Adversarial Agents
• Natural Language Processing: Question Answering

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: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) have demonstrated strong capabilities in language understanding and reasoning. However, their dependence on static training corpora makes them prone to factual errors and knowledge gaps. Retrieval-Augmented Generation (RAG) addresses this limitation by incorporating external knowledge sources, especially structured Knowledge Graphs (KGs), which provide explicit semantics and efficient retrieval. Existing KG-based RAG approaches, however, generally assume that anchor entities are accessible to initiate graph traversal, which limits their robustness in open world settings where accurate linking between the query and the entity is unreliable. To overcome this limitation, we propose AnchorRAG, a novel multi-agent collaboration framework for open-world RAG without the predefined anchor entities. Specifically, a predictor agent dynamically identifies candidate anchor entities by aligning user query terms with KG nodes and initializes independent retriever agents to conduct parallel multi-hop explorations from each candidate. Then a supervisor agent formulates the iterative retrieval strategy for these retriever agents and synthesizes the resulting knowledge paths to generate the final answer. This multi-agent collaboration framework improves retrieval robustness and mitigates the impact of ambiguous or erroneous anchors. Extensive experiments on four public benchmarks demonstrate that AnchorRAG significantly outperforms existing baselines and establishes new state-of-the-art results on the real-world question answering tasks.

Problem

• Addresses unreliable entity linking in open-world knowledge graph retrieval
• Overcomes dependency on predefined anchor entities for graph traversal
• Enhances robustness against ambiguous or erroneous anchor identification

Innovation

• Multi-agent collaboration framework for knowledge retrieval
• Dynamic anchor entity identification without predefined entities
• Parallel multi-hop exploration from candidate entities