<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yu-etal-2026-dimple">
    <titleInfo>
        <title>Dimple: Discrete Diffusion Parallel Generation for Large Multimodal Modal</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Runpeng</namePart>
        <namePart type="family">Yu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xinyin</namePart>
        <namePart type="family">Ma</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xinchao</namePart>
        <namePart type="family">Wang</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>This paper introduces Dimple and Dimple+, two discrete diffusion multimodal large language models (dMLLMs). Dimple is derived from a discrete diffusion large language model (dLLM) that initially lacks multimodal understanding and acquires this capability through a hybrid training strategy that begins with autoregressive learning and subsequently transitions to discrete diffusion training. In contrast, Dimple+ is initialized from an autoregressive multimodal large language model and attains parallel decoding ability via purely discrete diffusion training. Both models achieve performance on par with their autoregressive counterparts, while Dimple+ further sets a new state of the art among existing dMLLMs. To improve inference efficiency, we propose Confident Decoding, a method that dynamically adapts the number of tokens generated at each iteration. Experimental results demonstrate decoding speedups of 2\times–6\times with only marginal performance loss. Additionally, we show that the Prefilling technique, previously developed for autoregressive models, can be effectively extended to dMLLMs with bidirectional attention, yielding nearly lossless acceleration of 1.7\times–7\times. Finally, we present the Structure Prior method, which enables fine-grained control over output format and reasoning structure, a capability that is challenging to achieve in autoregressive models.</abstract>
    <identifier type="citekey">yu-etal-2026-dimple</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.824/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>19146</start>
            <end>19170</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Dimple: Discrete Diffusion Parallel Generation for Large Multimodal Modal
%A Yu, Runpeng
%A Ma, Xinyin
%A Wang, Xinchao
%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 yu-etal-2026-dimple
%X This paper introduces Dimple and Dimple+, two discrete diffusion multimodal large language models (dMLLMs). Dimple is derived from a discrete diffusion large language model (dLLM) that initially lacks multimodal understanding and acquires this capability through a hybrid training strategy that begins with autoregressive learning and subsequently transitions to discrete diffusion training. In contrast, Dimple+ is initialized from an autoregressive multimodal large language model and attains parallel decoding ability via purely discrete diffusion training. Both models achieve performance on par with their autoregressive counterparts, while Dimple+ further sets a new state of the art among existing dMLLMs. To improve inference efficiency, we propose Confident Decoding, a method that dynamically adapts the number of tokens generated at each iteration. Experimental results demonstrate decoding speedups of 2\times–6\times with only marginal performance loss. Additionally, we show that the Prefilling technique, previously developed for autoregressive models, can be effectively extended to dMLLMs with bidirectional attention, yielding nearly lossless acceleration of 1.7\times–7\times. Finally, we present the Structure Prior method, which enables fine-grained control over output format and reasoning structure, a capability that is challenging to achieve in autoregressive models.
%U https://aclanthology.org/2026.emnlp-main.824/
%P 19146-19170

Markdown (Informal)

[Dimple: Discrete Diffusion Parallel Generation for Large Multimodal Modal](https://aclanthology.org/2026.emnlp-main.824/) (Yu et al., EMNLP 2026)

• Dimple: Discrete Diffusion Parallel Generation for Large Multimodal Modal (Yu et al., EMNLP 2026)

ACL

• Runpeng Yu, Xinyin Ma, and Xinchao Wang. 2026. Dimple: Discrete Diffusion Parallel Generation for Large Multimodal Modal. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 19146–19170, Budapest, Hungary. Association for Computational Linguistics.