Debating Truth: Debate-driven Claim Verification with Multiple Large Language Model Agents

arXiv cs.CL / 4/6/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces DebateCV, a debate-driven claim verification framework that uses multiple LLM agents with two opposing “Debaters” and a “Moderator” to adjudicate evidence and reach accurate verdicts for complex claims.
It argues that single-agent claim verification can miss subtle errors when evidence is nuanced or multifaceted, motivating a structured adversarial debate setup to improve detection.
The authors identify a key limitation with zero-shot Moderators, noting they tend to produce biased or overly neutral judgments, and they state there are no existing datasets to train Moderators.
To address this, they propose Debate-SFT, a post-training approach that uses synthetic data to improve how Moderators weigh conflicting debate arguments.
Experimental results indicate the debate-based approach improves both accuracy across different evidence conditions and justification quality compared with non-debate state-of-the-art methods.

Abstract

State-of-the-art single-agent claim verification methods struggle with complex claims that require nuanced analysis of multifaceted evidence. Inspired by real-world professional fact-checkers, we propose \textbf{DebateCV}, the first debate-driven claim verification framework powered by multiple LLM agents. In DebateCV, two \textit{Debaters} argue opposing stances to surface subtle errors in single-agent assessments. A decisive \textit{Moderator} is then required to weigh the evidential strength of conflicting arguments to deliver an accurate verdict. Yet, zero-shot Moderators are biased toward neutral judgments, and no datasets exist for training them. To bridge this gap, we propose \textbf{Debate-SFT}, a post-training framework that leverages synthetic data to enhance agents' ability to effectively adjudicate debates for claim verification. Results show that our methods surpass state-of-the-art non-debate approaches in both accuracy (across various evidence conditions) and justification quality.