<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2025-amia">
    <titleInfo>
        <title>AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yuqi</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yuchun</namePart>
        <namePart type="family">Miao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zuchao</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Liang</namePart>
        <namePart type="family">Ding</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-11</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2025</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Christos</namePart>
            <namePart type="family">Christodoulopoulos</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Tanmoy</namePart>
            <namePart type="family">Chakraborty</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Carolyn</namePart>
            <namePart type="family">Rose</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Violet</namePart>
            <namePart type="family">Peng</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Suzhou, China</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-335-7</identifier>
    </relatedItem>
    <abstract>We introduce AMIA, a lightweight, inference-only defense for Large Vision–Language Models (LVLMs) that (1) Automatically Masks a small set of text-irrelevant image patches to disrupt adversarial perturbations, and (2) conducts joint Intention Analysis to uncover and mitigate hidden harmful intents before response generation. Without any retraining, AMIA improves defense success rates across diverse LVLMs and jailbreak benchmarks from an average of 52.4% to 81.7%, preserves general utility with only a 2% average accuracy drop, and incurs only modest inference overhead. Ablation confirms that both masking and intention analysis are essential for robust safety–utility trade-off. Our code will be released.</abstract>
    <identifier type="citekey">zhang-etal-2025-amia</identifier>
    <location>
        <url>https://aclanthology.org/2025.findings-emnlp.651/</url>
    </location>
    <part>
        <date>2025-11</date>
        <extent unit="page">
            <start>12189</start>
            <end>12199</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders
%A Zhang, Yuqi
%A Miao, Yuchun
%A Li, Zuchao
%A Ding, Liang
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Findings of the Association for Computational Linguistics: EMNLP 2025
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-335-7
%F zhang-etal-2025-amia
%X We introduce AMIA, a lightweight, inference-only defense for Large Vision–Language Models (LVLMs) that (1) Automatically Masks a small set of text-irrelevant image patches to disrupt adversarial perturbations, and (2) conducts joint Intention Analysis to uncover and mitigate hidden harmful intents before response generation. Without any retraining, AMIA improves defense success rates across diverse LVLMs and jailbreak benchmarks from an average of 52.4% to 81.7%, preserves general utility with only a 2% average accuracy drop, and incurs only modest inference overhead. Ablation confirms that both masking and intention analysis are essential for robust safety–utility trade-off. Our code will be released.
%U https://aclanthology.org/2025.findings-emnlp.651/
%P 12189-12199

Markdown (Informal)

[AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders](https://aclanthology.org/2025.findings-emnlp.651/) (Zhang et al., Findings 2025)

• AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders (Zhang et al., Findings 2025)

ACL

• Yuqi Zhang, Yuchun Miao, Zuchao Li, and Liang Ding. 2025. AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 12189–12199, Suzhou, China. Association for Computational Linguistics.