@inproceedings{cha-chae-2026-attention,
title = "Attention-Aware Remasking: A Training-Free Approach for Enhancing Visual Fidelity in Multimodal Diffusion Language Models",
author = "Cha, Yechan and
Chae, Dong-Kyu",
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.765/",
pages = "16137--16151",
ISBN = "979-8-89176-462-0",
abstract = "Multimodal diffusion Large Language Models (dMLLMs) emerge as a compelling alternative to traditional autoregressive paradigms. But they still suffer from hallucinations driven by an inherent language bias. Through an analysis of the dMLLM inference dynamics, we observe that visual tokens frequently receive disproportionately low attention scores. We identify that the conventional low-confidence remasking strategy redundantly leverages logit-based probabilities during the token selection process, thereby exacerbating this language bias. To address this vulnerability, we propose Attention-Aware Remasking, a training-free, plug-and-play strategy that incorporates newly defined visual confidence scores to select the final output tokens, effectively promoting visual grounding. To enhance the logical integrity and inference efficiency of dLLMs, we further propose clue-based answering and echo-filling. Evaluations on representative Visual-Language Natural Language Explanation (VL-NLE) tasks, including VQA-X and e-SNLI-VE, demonstrate that our method achieves the highest BERTScore on both VQA-X and e-SNLI-VE, and the best answer accuracy on VQA-X among the compared methods. Our study reveals that the widely adopted logit-based remasking is not necessarily the optimal approach, highlighting promising new avenues for visually grounded inference in dMLLMs. Our code is available at: https://github.com/yechancha/Attention-aware-Remasking."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cha-chae-2026-attention">
<titleInfo>
<title>Attention-Aware Remasking: A Training-Free Approach for Enhancing Visual Fidelity in Multimodal Diffusion Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yechan</namePart>
<namePart type="family">Cha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dong-Kyu</namePart>
<namePart type="family">Chae</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 diffusion Large Language Models (dMLLMs) emerge as a compelling alternative to traditional autoregressive paradigms. But they still suffer from hallucinations driven by an inherent language bias. Through an analysis of the dMLLM inference dynamics, we observe that visual tokens frequently receive disproportionately low attention scores. We identify that the conventional low-confidence remasking strategy redundantly leverages logit-based probabilities during the token selection process, thereby exacerbating this language bias. To address this vulnerability, we propose Attention-Aware Remasking, a training-free, plug-and-play strategy that incorporates newly defined visual confidence scores to select the final output tokens, effectively promoting visual grounding. To enhance the logical integrity and inference efficiency of dLLMs, we further propose clue-based answering and echo-filling. Evaluations on representative Visual-Language Natural Language Explanation (VL-NLE) tasks, including VQA-X and e-SNLI-VE, demonstrate that our method achieves the highest BERTScore on both VQA-X and e-SNLI-VE, and the best answer accuracy on VQA-X among the compared methods. Our study reveals that the widely adopted logit-based remasking is not necessarily the optimal approach, highlighting promising new avenues for visually grounded inference in dMLLMs. Our code is available at: https://github.com/yechancha/Attention-aware-Remasking.</abstract>
<identifier type="citekey">cha-chae-2026-attention</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.765/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16137</start>
<end>16151</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Attention-Aware Remasking: A Training-Free Approach for Enhancing Visual Fidelity in Multimodal Diffusion Language Models
%A Cha, Yechan
%A Chae, Dong-Kyu
%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 cha-chae-2026-attention
%X Multimodal diffusion Large Language Models (dMLLMs) emerge as a compelling alternative to traditional autoregressive paradigms. But they still suffer from hallucinations driven by an inherent language bias. Through an analysis of the dMLLM inference dynamics, we observe that visual tokens frequently receive disproportionately low attention scores. We identify that the conventional low-confidence remasking strategy redundantly leverages logit-based probabilities during the token selection process, thereby exacerbating this language bias. To address this vulnerability, we propose Attention-Aware Remasking, a training-free, plug-and-play strategy that incorporates newly defined visual confidence scores to select the final output tokens, effectively promoting visual grounding. To enhance the logical integrity and inference efficiency of dLLMs, we further propose clue-based answering and echo-filling. Evaluations on representative Visual-Language Natural Language Explanation (VL-NLE) tasks, including VQA-X and e-SNLI-VE, demonstrate that our method achieves the highest BERTScore on both VQA-X and e-SNLI-VE, and the best answer accuracy on VQA-X among the compared methods. Our study reveals that the widely adopted logit-based remasking is not necessarily the optimal approach, highlighting promising new avenues for visually grounded inference in dMLLMs. Our code is available at: https://github.com/yechancha/Attention-aware-Remasking.
%U https://aclanthology.org/2026.findings-emnlp.765/
%P 16137-16151
Markdown (Informal)
[Attention-Aware Remasking: A Training-Free Approach for Enhancing Visual Fidelity in Multimodal Diffusion Language Models](https://aclanthology.org/2026.findings-emnlp.765/) (Cha & Chae, Findings 2026)
ACL