VERDICT: Training-Free Step-Wise Verification of Multimodal Reasoning via Disagreement-Aware Consensus

2026-08-11

AI Summary

This work addresses the challenge that reasoning chains generated by multimodal large language models are often derailed by subtle errors, while existing verification approaches either rely on costly annotated data or overlook disagreement signals among verification sources. The authors propose a training-free, domain-agnostic stepwise verification method that models the scores from multiple frozen multimodal verifiers as a coordination game, explicitly capturing cross-modal disagreement and deriving a consensus score to effectively filter and rank reasoning steps. This approach is the first to operationalize cross-modal verification disagreement as an actionable signal, introducing a closed-form coupled scoring mechanism and a disagreement-aware thresholding strategy. Evaluated across six benchmarks, the method achieves an average improvement of +5.95%, matching the performance of heavily supervised, domain-specific verifiers and demonstrating the efficacy of training-free cross-modal consensus.

Abstract

Multimodal large language models often generate reasoning chains containing subtle errors that lead to incorrect answers. Current verification approaches have notable limitations. Existing approaches either require expensive labelled supervision with inconsistent cross-task performance or aggregate scores from multiple sources by simple aggregations, missing a key insight: when these scores disagree, that disagreement itself carries important information about whether a reasoning step is truly valid or not. We formalise this as a coupled scoring problem among disparate, frozen verifiers, interpretable as a coordination game with a unique closed-form equilibrium where agreement signals valid steps while disagreement reveals instability. Towards this end, we propose a training-free domain-agnostic step-wise verification approach we call VERDICT: VERification via Disagreement-Informed Coupled Thresholding. To our knowledge, VERDICT is the first training-free verifier that makes the structure of cross-modal disagreement explicit and actionable. It computes consensus scores through a closed-form solution, enabling both disagreement-aware filtering and stability-conscious ranking of reasoning steps. Evaluated across six benchmarks, \method consistently improves over the base model by up to +5.95%, and performs competitively with domain-specific critics that demand extensive supervision, demonstrating that cross-modal agreement provides robust verification signals without task-specific adaptation and Training-Free Verification

Problem

• multimodal reasoning
• verification
• disagreement-aware
• training-free
• reasoning errors

Innovation

• training-free verification
• disagreement-aware consensus
• multimodal reasoning
• coupled scoring
• closed-form equilibrium