<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="derner-batistic-2025-beyond">
    <titleInfo>
        <title>Beyond Words: Multilingual and Multimodal Red Teaming of MLLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Erik</namePart>
        <namePart type="family">Derner</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kristina</namePart>
        <namePart type="family">Batistič</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-08</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the The First Workshop on LLM Security (LLMSEC)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Leon</namePart>
            <namePart type="family">Derczynski</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jekaterina</namePart>
            <namePart type="family">Novikova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Muhao</namePart>
            <namePart type="family">Chen</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">979-8-89176-279-4</identifier>
    </relatedItem>
    <abstract>Multimodal large language models (MLLMs) are increasingly deployed in real-world applications, yet their safety remains underexplored, particularly in multilingual and visual contexts. In this work, we present a systematic red teaming framework to evaluate MLLM safeguards using adversarial prompts translated into seven languages and delivered via four input modalities: plain text, jailbreak prompt + text, text rendered as an image, and jailbreak prompt + text rendered as an image. We find that rendering prompts as images increases attack success rates and reduces refusal rates, with the effect most pronounced in lower-resource languages such as Slovenian, Czech, and Valencian. Our results suggest that vision-based multilingual attacks expose a persistent gap in current alignment strategies, highlighting the need for robust multilingual and multimodal MLLM safety evaluation and mitigation of these risks. We make our code and data available.</abstract>
    <identifier type="citekey">derner-batistic-2025-beyond</identifier>
    <location>
        <url>https://aclanthology.org/2025.llmsec-1.15/</url>
    </location>
    <part>
        <date>2025-08</date>
        <extent unit="page">
            <start>198</start>
            <end>206</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Beyond Words: Multilingual and Multimodal Red Teaming of MLLMs
%A Derner, Erik
%A Batistič, Kristina
%Y Derczynski, Leon
%Y Novikova, Jekaterina
%Y Chen, Muhao
%S Proceedings of the The First Workshop on LLM Security (LLMSEC)
%D 2025
%8 August
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-279-4
%F derner-batistic-2025-beyond
%X Multimodal large language models (MLLMs) are increasingly deployed in real-world applications, yet their safety remains underexplored, particularly in multilingual and visual contexts. In this work, we present a systematic red teaming framework to evaluate MLLM safeguards using adversarial prompts translated into seven languages and delivered via four input modalities: plain text, jailbreak prompt + text, text rendered as an image, and jailbreak prompt + text rendered as an image. We find that rendering prompts as images increases attack success rates and reduces refusal rates, with the effect most pronounced in lower-resource languages such as Slovenian, Czech, and Valencian. Our results suggest that vision-based multilingual attacks expose a persistent gap in current alignment strategies, highlighting the need for robust multilingual and multimodal MLLM safety evaluation and mitigation of these risks. We make our code and data available.
%U https://aclanthology.org/2025.llmsec-1.15/
%P 198-206

Markdown (Informal)

[Beyond Words: Multilingual and Multimodal Red Teaming of MLLMs](https://aclanthology.org/2025.llmsec-1.15/) (Derner & Batistič, LLMSEC 2025)

• Beyond Words: Multilingual and Multimodal Red Teaming of MLLMs (Derner & Batistič, LLMSEC 2025)

ACL

• Erik Derner and Kristina Batistič. 2025. Beyond Words: Multilingual and Multimodal Red Teaming of MLLMs. In Proceedings of the The First Workshop on LLM Security (LLMSEC), pages 198–206, Vienna, Austria. Association for Computational Linguistics.