@inproceedings{meng-etal-2026-visual,
title = "Visual Representation Regularization for Multimodal Hallucination Mitigation",
author = "Meng, Yifan and
Wang, Enguang and
Cao, Xusheng and
Yang, Fei and
Liu, Xialei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1561/",
pages = "33253--33267",
ISBN = "979-8-89176-462-0",
abstract = "Multimodal large language models (MLLMs) have achieved remarkable progress in multimodal reasoning, yet they still suffer from hallucinations that undermine their reliability. Existing inference-time mitigation methods mainly focus on optimizing decoding strategies or analyzing internal attention mechanisms, while the role of visual representations remains underexplored. In this work, we reveal that visual representations exhibit information dispersion during layer-wise evolution, manifested as a progressive deviation from the initial visual subspace and increasing alignment with the textual subspace. Based on this insight, we propose Visual Representation Regularization (VRR), a training-free method that mitigates hallucinations by regularizing visual representations toward their initial states. Extensive experiments demonstrate that VRR consistently reduces hallucinations and is compatible with existing mitigation methods. Our code is available at https://github.com/Jiarker/VRR."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="meng-etal-2026-visual">
<titleInfo>
<title>Visual Representation Regularization for Multimodal Hallucination Mitigation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="family">Meng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enguang</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xusheng</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fei</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xialei</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Multimodal large language models (MLLMs) have achieved remarkable progress in multimodal reasoning, yet they still suffer from hallucinations that undermine their reliability. Existing inference-time mitigation methods mainly focus on optimizing decoding strategies or analyzing internal attention mechanisms, while the role of visual representations remains underexplored. In this work, we reveal that visual representations exhibit information dispersion during layer-wise evolution, manifested as a progressive deviation from the initial visual subspace and increasing alignment with the textual subspace. Based on this insight, we propose Visual Representation Regularization (VRR), a training-free method that mitigates hallucinations by regularizing visual representations toward their initial states. Extensive experiments demonstrate that VRR consistently reduces hallucinations and is compatible with existing mitigation methods. Our code is available at https://github.com/Jiarker/VRR.</abstract>
<identifier type="citekey">meng-etal-2026-visual</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1561/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>33253</start>
<end>33267</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Visual Representation Regularization for Multimodal Hallucination Mitigation
%A Meng, Yifan
%A Wang, Enguang
%A Cao, Xusheng
%A Yang, Fei
%A Liu, Xialei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F meng-etal-2026-visual
%X Multimodal large language models (MLLMs) have achieved remarkable progress in multimodal reasoning, yet they still suffer from hallucinations that undermine their reliability. Existing inference-time mitigation methods mainly focus on optimizing decoding strategies or analyzing internal attention mechanisms, while the role of visual representations remains underexplored. In this work, we reveal that visual representations exhibit information dispersion during layer-wise evolution, manifested as a progressive deviation from the initial visual subspace and increasing alignment with the textual subspace. Based on this insight, we propose Visual Representation Regularization (VRR), a training-free method that mitigates hallucinations by regularizing visual representations toward their initial states. Extensive experiments demonstrate that VRR consistently reduces hallucinations and is compatible with existing mitigation methods. Our code is available at https://github.com/Jiarker/VRR.
%U https://aclanthology.org/2026.findings-emnlp.1561/
%P 33253-33267
Markdown (Informal)
[Visual Representation Regularization for Multimodal Hallucination Mitigation](https://aclanthology.org/2026.findings-emnlp.1561/) (Meng et al., Findings 2026)
ACL