Multi-Sourced, Multi-Agent Evidence Retrieval for Fact-Checking

2026-02-27

AI Summary

This work addresses the limitations of existing fact-checking methods in modeling multi-hop semantic relationships during evidence retrieval, which often leads to poor generalization and inaccurate identification of factual associations. To overcome these challenges, the authors propose a multi-agent fact-checking framework that integrates open knowledge graphs with web content. The approach uniquely treats the knowledge graph as the primary evidence source and leverages large language models (LLMs) to jointly retrieve structured and unstructured evidence. A Markov decision process (MDP)-based multi-agent retrieval mechanism is introduced to navigate complex evidence paths, while prompt optimization techniques enhance the LLM’s reasoning and decision-making capabilities. Experimental results demonstrate that this method significantly improves evidence relevance and fact-checking accuracy, effectively capturing fine-grained factual relationships embedded in complex claims.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Knowledge Representation and Reasoning: Reasoning with Beliefs
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• 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
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Misinformation spreading over the Internet poses a significant threat to both societies and individuals, necessitating robust and scalable fact-checking that relies on retrieving accurate and trustworthy evidence. Previous methods rely on semantic and social-contextual patterns learned from training data, which limits their generalization to new data distributions. Recently, Retrieval Augmented Generation (RAG) based methods have been proposed to utilize the reasoning capability of LLMs with retrieved grounding evidence documents. However, these methods largely rely on textual similarity for evidence retrieval and struggle to retrieve evidence that captures multi-hop semantic relations within rich document contents. These limitations lead to overlooking subtle factual correlations between the evidence and the claims to be fact-checked during evidence retrieval, thus causing inaccurate veracity predictions. To address these issues, we propose WKGFC, which exploits authorized open knowledge graph as a core resource of evidence. LLM-enabled retrieval is designed to assess the claims and retrieve the most relevant knowledge subgraphs, forming structured evidence for fact verification. To augment the knowledge graph evidence, we retrieve web contents for completion. The above process is implemented as an automatic Markov Decision Process (MDP): A reasoning LLM agent decides what actions to take according to the current evidence and the claims. To adapt the MDP for fact-checking, we use prompt optimization to fine-tune the agentic LLM.

Problem

• fact-checking
• evidence retrieval
• multi-hop reasoning
• misinformation
• knowledge graph

Innovation

• Knowledge Graph
• Retrieval-Augmented Generation
• Multi-Agent Retrieval
• Markov Decision Process
• Fact-Checking