@inproceedings{chang-etal-2025-focus,
title = "Focus on What Matters: Enhancing Medical Vision-Language Models with Automatic Attention Alignment Tuning",
author = "Chang, Aofei and
Huang, Le and
Boyd, Alex James and
Bhatia, Parminder and
Kass-Hout, Taha and
Xiao, Cao and
Ma, Fenglong",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.460/",
doi = "10.18653/v1/2025.acl-long.460",
pages = "9357--9372",
ISBN = "979-8-89176-251-0",
abstract = "Medical Large Vision-Language Models (Med-LVLMs) often exhibit suboptimal attention distribution on visual inputs, leading to hallucinated or inaccurate outputs. Existing methods primarily rely on inference-time interventions, which are limited in attention adaptation or require additional supervision. To address this, we propose A$^3$Tune, a novel fine-tuning framework for Automatic Attention Alignment Tuning. ATune leverages zero-shot weak labels from SAM, refines them into prompt-aware labels using BioMedCLIP, and then selectively modifies visually-critical attention heads to improve alignment while minimizing interference. Additionally, we introduce a A$^3$MoE module, enabling adaptive parameter selection for attention tuning across diverse prompts and images. Extensive experiments on medical VQA and report generation benchmarks show that A$^3$Tune outperforms state-of-the-art baselines, achieving enhanced attention distributions and performance in Med-LVLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chang-etal-2025-focus">
<titleInfo>
<title>Focus on What Matters: Enhancing Medical Vision-Language Models with Automatic Attention Alignment Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aofei</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Le</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alex</namePart>
<namePart type="given">James</namePart>
<namePart type="family">Boyd</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Parminder</namePart>
<namePart type="family">Bhatia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taha</namePart>
<namePart type="family">Kass-Hout</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cao</namePart>
<namePart type="family">Xiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fenglong</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Medical Large Vision-Language Models (Med-LVLMs) often exhibit suboptimal attention distribution on visual inputs, leading to hallucinated or inaccurate outputs. Existing methods primarily rely on inference-time interventions, which are limited in attention adaptation or require additional supervision. To address this, we propose A³Tune, a novel fine-tuning framework for Automatic Attention Alignment Tuning. ATune leverages zero-shot weak labels from SAM, refines them into prompt-aware labels using BioMedCLIP, and then selectively modifies visually-critical attention heads to improve alignment while minimizing interference. Additionally, we introduce a A³MoE module, enabling adaptive parameter selection for attention tuning across diverse prompts and images. Extensive experiments on medical VQA and report generation benchmarks show that A³Tune outperforms state-of-the-art baselines, achieving enhanced attention distributions and performance in Med-LVLMs.</abstract>
<identifier type="citekey">chang-etal-2025-focus</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.460</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.460/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>9357</start>
<end>9372</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Focus on What Matters: Enhancing Medical Vision-Language Models with Automatic Attention Alignment Tuning
%A Chang, Aofei
%A Huang, Le
%A Boyd, Alex James
%A Bhatia, Parminder
%A Kass-Hout, Taha
%A Xiao, Cao
%A Ma, Fenglong
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F chang-etal-2025-focus
%X Medical Large Vision-Language Models (Med-LVLMs) often exhibit suboptimal attention distribution on visual inputs, leading to hallucinated or inaccurate outputs. Existing methods primarily rely on inference-time interventions, which are limited in attention adaptation or require additional supervision. To address this, we propose A³Tune, a novel fine-tuning framework for Automatic Attention Alignment Tuning. ATune leverages zero-shot weak labels from SAM, refines them into prompt-aware labels using BioMedCLIP, and then selectively modifies visually-critical attention heads to improve alignment while minimizing interference. Additionally, we introduce a A³MoE module, enabling adaptive parameter selection for attention tuning across diverse prompts and images. Extensive experiments on medical VQA and report generation benchmarks show that A³Tune outperforms state-of-the-art baselines, achieving enhanced attention distributions and performance in Med-LVLMs.
%R 10.18653/v1/2025.acl-long.460
%U https://aclanthology.org/2025.acl-long.460/
%U https://doi.org/10.18653/v1/2025.acl-long.460
%P 9357-9372
Markdown (Informal)
[Focus on What Matters: Enhancing Medical Vision-Language Models with Automatic Attention Alignment Tuning](https://aclanthology.org/2025.acl-long.460/) (Chang et al., ACL 2025)
ACL