<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liyan-etal-2026-mitigating">
    <titleInfo>
        <title>Mitigating Object Hallucinations in Vision-Language Models via Head-Level Attribution and Counteractive Pruning</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Wu</namePart>
        <namePart type="family">Liyan</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xin</namePart>
        <namePart type="family">Luo</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Fan</namePart>
        <namePart type="family">Wu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yingen</namePart>
        <namePart type="family">Liu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name>
        <namePart>Wuxiaobin</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>Large vision-language models (LVLMs) excel at multimodal tasks but often generate instance-level object hallucinations, describing nonexistent objects. Yet most mitigation strategies rely on output-level or coarse-grained interventions, limiting their ability to identify and selectively regulate attention heads associated with faithful versus hallucinatory generation, often at the cost of descriptive coverage. To address this limitation, we propose HACP, a training-free framework that integrates head-level semantic attribution with instance-adaptive counteractive pruning. HACP introduces InfoSpectralScore to summarize spectral changes in head-perturbed representations and identify heads associated with hallucination or grounded generation, which are then selectively suppressed or reinforced during decoding. Across multiple LVLMs and benchmarks, HACP consistently reduces hallucinations while preserving descriptive coverage; on a 1000-image multi-label COCO subset, it lowers CHAIR-s/i from 66.7/18.0 to 43.4/8.6 and improves F1 from 73.0 to 77.0. These findings suggest that head-level attribution provides an actionable signal for fine-grained inference-time hallucination control in LVLMs, while helping preserve descriptive coverage. The official project repository for HACP is available at https://github.com/S1yang/HACP.</abstract>
    <identifier type="citekey">liyan-etal-2026-mitigating</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1505/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>32040</start>
            <end>32060</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Mitigating Object Hallucinations in Vision-Language Models via Head-Level Attribution and Counteractive Pruning
%A Liyan, Wu
%A Luo, Xin
%A Wu, Fan
%A Liu, Yingen
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%A Wuxiaobin
%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 liyan-etal-2026-mitigating
%X Large vision-language models (LVLMs) excel at multimodal tasks but often generate instance-level object hallucinations, describing nonexistent objects. Yet most mitigation strategies rely on output-level or coarse-grained interventions, limiting their ability to identify and selectively regulate attention heads associated with faithful versus hallucinatory generation, often at the cost of descriptive coverage. To address this limitation, we propose HACP, a training-free framework that integrates head-level semantic attribution with instance-adaptive counteractive pruning. HACP introduces InfoSpectralScore to summarize spectral changes in head-perturbed representations and identify heads associated with hallucination or grounded generation, which are then selectively suppressed or reinforced during decoding. Across multiple LVLMs and benchmarks, HACP consistently reduces hallucinations while preserving descriptive coverage; on a 1000-image multi-label COCO subset, it lowers CHAIR-s/i from 66.7/18.0 to 43.4/8.6 and improves F1 from 73.0 to 77.0. These findings suggest that head-level attribution provides an actionable signal for fine-grained inference-time hallucination control in LVLMs, while helping preserve descriptive coverage. The official project repository for HACP is available at https://github.com/S1yang/HACP.
%U https://aclanthology.org/2026.findings-emnlp.1505/
%P 32040-32060

Markdown (Informal)

[Mitigating Object Hallucinations in Vision-Language Models via Head-Level Attribution and Counteractive Pruning](https://aclanthology.org/2026.findings-emnlp.1505/) (Liyan et al., Findings 2026)

• Mitigating Object Hallucinations in Vision-Language Models via Head-Level Attribution and Counteractive Pruning (Liyan et al., Findings 2026)

ACL

• Wu Liyan, Xin Luo, Fan Wu, Yingen Liu, and Wuxiaobin. 2026. Mitigating Object Hallucinations in Vision-Language Models via Head-Level Attribution and Counteractive Pruning. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 32040–32060, Budapest, Hungary. Association for Computational Linguistics.