<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gou-etal-2025-sure">
    <titleInfo>
        <title>SURE: Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yuxin</namePart>
        <namePart type="family">Gou</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xiaoning</namePart>
        <namePart type="family">Dong</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qin</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Shishen</namePart>
        <namePart type="family">Gu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Richang</namePart>
        <namePart type="family">Hong</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wenbo</namePart>
        <namePart type="family">Hu</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>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-332-6</identifier>
    </relatedItem>
    <abstract>Multimodal large language models (MLLMs) demonstrate impressive capabilities by integrating visual and textual information. However, the incorporation of visual modalities also introduces new and complex safety risks, rendering even the most advanced models vulnerable to sophisticated jailbreak attacks. This paper first analyzes the impact of inserting safety reasoning prompt on various aspects of the model. We find that this external method can help the model resist jailbreak attacks to some extent, but the model still fails to distinguish specific semantic scenarios, resulting in a significantly increased refusal rate for benign queries. Inspired by this, we propose a novel training framework, SURE (Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models), designed to help models internalize chain-of-thought-based safety decision-making capabilities. Extensive experiments demonstrate that SURE significantly improves model safety while effectively avoiding over-defense, achieving a good balance between safety and generality. Finally, we create a large-scale multimodal safety reasoning dataset, MLLM-SCoT-Plus, to facilitate research on safety alignment in multimodal models.Our code and the dataset are publicly available at https://github.com/hfutml/SURE.</abstract>
    <identifier type="citekey">gou-etal-2025-sure</identifier>
    <identifier type="doi">10.18653/v1/2025.emnlp-main.384</identifier>
    <location>
        <url>https://aclanthology.org/2025.emnlp-main.384/</url>
    </location>
    <part>
        <date>2025-11</date>
        <extent unit="page">
            <start>7563</start>
            <end>7604</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T SURE: Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models
%A Gou, Yuxin
%A Dong, Xiaoning
%A Li, Qin
%A Gu, Shishen
%A Hong, Richang
%A Hu, Wenbo
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-332-6
%F gou-etal-2025-sure
%X Multimodal large language models (MLLMs) demonstrate impressive capabilities by integrating visual and textual information. However, the incorporation of visual modalities also introduces new and complex safety risks, rendering even the most advanced models vulnerable to sophisticated jailbreak attacks. This paper first analyzes the impact of inserting safety reasoning prompt on various aspects of the model. We find that this external method can help the model resist jailbreak attacks to some extent, but the model still fails to distinguish specific semantic scenarios, resulting in a significantly increased refusal rate for benign queries. Inspired by this, we propose a novel training framework, SURE (Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models), designed to help models internalize chain-of-thought-based safety decision-making capabilities. Extensive experiments demonstrate that SURE significantly improves model safety while effectively avoiding over-defense, achieving a good balance between safety and generality. Finally, we create a large-scale multimodal safety reasoning dataset, MLLM-SCoT-Plus, to facilitate research on safety alignment in multimodal models.Our code and the dataset are publicly available at https://github.com/hfutml/SURE.
%R 10.18653/v1/2025.emnlp-main.384
%U https://aclanthology.org/2025.emnlp-main.384/
%U https://doi.org/10.18653/v1/2025.emnlp-main.384
%P 7563-7604

Markdown (Informal)

[SURE: Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models](https://aclanthology.org/2025.emnlp-main.384/) (Gou et al., EMNLP 2025)

• SURE: Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models (Gou et al., EMNLP 2025)

ACL

• Yuxin Gou, Xiaoning Dong, Qin Li, Shishen Gu, Richang Hong, and Wenbo Hu. 2025. SURE: Safety Understanding and Reasoning Enhancement for Multimodal Large Language Models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 7563–7604, Suzhou, China. Association for Computational Linguistics.