<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhao-etal-2026-context">
    <titleInfo>
        <title>Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yikai</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qiyan</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiaquan</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xiaofeng</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xiaosong</namePart>
        <namePart type="family">Yuan</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Pengzhou</namePart>
        <namePart type="family">Cheng</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Diffusion multimodal large language models (dMLLMs) frequently produce long-form outputs marred by semantic drift and repetition, with quality generally degrading as output length increases. We identify two structural deficiencies in existing decoding methods as primary drivers of these failures: confidence-based scoring ignores decoded-neighbor support, and block partitioning prevents access to high-readiness semantic anchors. Together, these deficiencies cause tokens to be committed before their local context is sufficiently established. We propose CACD (Context-Aware Cluster Decoding), a training-free decoding method that scores each masked position using a multiplicative combination of softmax confidence and neighbor proximity. This approach prioritizes contextually ready tokens over isolated candidates while suppressing low-confidence positional noise. CACD operates without block partitioning, keeping high-readiness anchors globally accessible. It further applies architecture-aware calibration to handle confidence heterogeneity induced by diverse visual integration strategies. Experiments on three dMLLMs across four benchmarks demonstrate consistent quality gains and hallucination reduction over the original decoding methods, with larger gains in several longer-generation settings. These results highlight the importance of neighbor support and visual integration strategies for the future design of dMLLM decoding methods. Our code is openly available at https://github.com/zhaoyk-sysu/CACD-dMLLM.</abstract>
    <identifier type="citekey">zhao-etal-2026-context</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2072/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>47182</start>
            <end>47200</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs
%A Zhao, Yikai
%A Zhao, Qiyan
%A Zhang, Jiaquan
%A Zhang, Xiaofeng
%A Yuan, Xiaosong
%A Cheng, Pengzhou
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F zhao-etal-2026-context
%X Diffusion multimodal large language models (dMLLMs) frequently produce long-form outputs marred by semantic drift and repetition, with quality generally degrading as output length increases. We identify two structural deficiencies in existing decoding methods as primary drivers of these failures: confidence-based scoring ignores decoded-neighbor support, and block partitioning prevents access to high-readiness semantic anchors. Together, these deficiencies cause tokens to be committed before their local context is sufficiently established. We propose CACD (Context-Aware Cluster Decoding), a training-free decoding method that scores each masked position using a multiplicative combination of softmax confidence and neighbor proximity. This approach prioritizes contextually ready tokens over isolated candidates while suppressing low-confidence positional noise. CACD operates without block partitioning, keeping high-readiness anchors globally accessible. It further applies architecture-aware calibration to handle confidence heterogeneity induced by diverse visual integration strategies. Experiments on three dMLLMs across four benchmarks demonstrate consistent quality gains and hallucination reduction over the original decoding methods, with larger gains in several longer-generation settings. These results highlight the importance of neighbor support and visual integration strategies for the future design of dMLLM decoding methods. Our code is openly available at https://github.com/zhaoyk-sysu/CACD-dMLLM.
%U https://aclanthology.org/2026.emnlp-main.2072/
%P 47182-47200

Markdown (Informal)

[Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs](https://aclanthology.org/2026.emnlp-main.2072/) (Zhao et al., EMNLP 2026)

• Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs (Zhao et al., EMNLP 2026)

ACL

• Yikai Zhao, Qiyan Zhao, Jiaquan Zhang, Xiaofeng Zhang, Xiaosong Yuan, and Pengzhou Cheng. 2026. Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 47182–47200, Budapest, Hungary. Association for Computational Linguistics.