Reflect to Inform: Boosting Multimodal Reasoning via Information-Gain-Driven Verification

Read full text Summarise with VIRAT
Authors: Shuai Lv, Chang Liu, Feng Tang, Rui Chen, Yu-Jie Yuan, Aojun Zhou, Kui Zhang, Xi Yang, Yangqiu Song
Cited by: 0
References: 63

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Multimodal Large Language Models (MLLMs) achieve strong multimodal reasoning performance, yet we identify a recurring failure mode in long-form generation: as outputs grow longer, models progressively drift away from image evidence and fall back on textual priors, resulting in ungrounded reasoning and hallucinations.Interestingly, Based on attention analysis, we find that MLLMs have a latent capability for late-stage visual verification that is present but not consistently activated.Motivated by this observation, we propose Visual Re-Examination (VRE), a self-evolving training framework that enables MLLMs to autonomously perform visual introspection during reasoning without additional visual inputs.Rather than distilling visual capabilities from a stronger teacher, VRE promotes iterative self-improvement by leveraging the model itself to generate reflection traces, making visual information actionable through information gain.Extensive experiments across diverse multimodal benchmarks demonstrate that VRE consistently improves reasoning accuracy and perceptual reliability, while substantially reducing hallucinations, especially in longchain settings.Code is available at https: //github.com/Xiaobu-USTC/VRE.

Fields of Study & Subjects

Multimodal Machine Learning Applications Topic Modeling Generative Adversarial Networks and Image Synthesis

Publication details

Journal / source
Conference on Empirical Methods in Natural Language Processing (EMNLP)
Pages
27318–27342
Publication date
1 January 2026
Language
EN