SP
BravenNow
Multi-Sourced, Multi-Agent Evidence Retrieval for Fact-Checking
| USA | technology | ✓ Verified - arxiv.org

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

#fact-checking #misinformation #knowledge graphs #LLM #WKGFC #retrieval augmented generation #Markov Decision Process #semantic relations

📌 Key Takeaways

  • Researchers developed WKGFC method for improved fact-checking using knowledge graphs
  • The approach addresses limitations in current methods that struggle with generalization to new data
  • The system uses LLM agents to retrieve relevant knowledge subgraphs and web content
  • Prompt optimization enhances the LLM's ability to capture multi-hop semantic relations
  • The research was published on arXiv on February 27, 2026

📖 Full Retelling

A team of researchers led by Shuzhi Gong from the University of Melbourne has developed a novel fact-checking method called WKGFC, published on arXiv on February 27, 2026, to combat the growing threat of misinformation spreading across the internet. The research addresses critical limitations in current fact-checking approaches that rely on semantic and social-contextual patterns learned from training data, which struggle to generalize to new data distributions. The proposed WKGFC system represents a significant advancement by utilizing authorized open knowledge graphs as a core resource of evidence, enabling more accurate and scalable fact verification in an era where false information increasingly threatens both societies and individuals. The researchers implemented their approach as an automatic Markov Decision Process where a reasoning LLM agent evaluates claims and retrieves the most relevant knowledge subgraphs, forming structured evidence for fact verification while also retrieving web contents to augment the knowledge graph evidence. To optimize this process, the team used prompt optimization to fine-tune the agentic LLM, allowing it to better capture multi-hop semantic relations within rich document contents that previous methods had overlooked.

🏷️ Themes

Artificial Intelligence, Fact-Checking, Knowledge Graphs, Misinformation

📚 Related People & Topics

Markov decision process

Mathematical model for sequential decision making under uncertainty

A Markov decision process (MDP) is a mathematical model for sequential decision making when outcomes are uncertain. It is a type of stochastic decision process, and is often solved using the methods of stochastic dynamic programming. Originating from operations research in the 1950s, MDPs have since...

View Profile → Wikipedia ↗

Large language model

Type of machine learning model

A large language model (LLM) is a language model trained with self-supervised machine learning on a vast amount of text, designed for natural language processing tasks, especially language generation. The largest and most capable LLMs are generative pre-trained transformers (GPTs) that provide the c...

View Profile → Wikipedia ↗

Entity Intersection Graph

No entity connections available yet for this article.

Mentioned Entities

Markov decision process

Mathematical model for sequential decision making under uncertainty

Large language model

Type of machine learning model
}
Original Source
--> Computer Science > Artificial Intelligence arXiv:2603.00267 [Submitted on 27 Feb 2026] Title: Multi-Sourced, Multi-Agent Evidence Retrieval for Fact-Checking Authors: Shuzhi Gong , Richard O. Sinnott , Jianzhong Qi , Cecile Paris , Preslav Nakov , Zhuohan Xie View a PDF of the paper titled Multi-Sourced, Multi-Agent Evidence Retrieval for Fact-Checking, by Shuzhi Gong and Richard O. Sinnott and Jianzhong Qi and Cecile Paris and Preslav Nakov and Zhuohan Xie View PDF HTML 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 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 : 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 optimiza...
Read full article at source

Source

arxiv.org

More from USA

News from Other Countries

🇬🇧 United Kingdom

🇺🇦 Ukraine