<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="fu-etal-2026-vismind">
    <titleInfo>
        <title>VisMind: Stateful Synthetic Dialogue for Visual Mid-Training</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Xinlong</namePart>
        <namePart type="family">Fu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zihao</namePart>
        <namePart type="family">Zhou</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qiufeng</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>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>The mid-training phase has emerged as a critical bridge for enhancing the reasoning capabilities of Vision-Language Models, yet its effectiveness is constrained by the scarcity of high-quality visual reasoning data. While data synthesis offers a promising alternative, conventional stateless synthesis paradigms produce shallow trajectories that quickly succumb to utility plateaus. To overcome this, we propose VisMind, a novel multi-agent collaborative framework that generates deep-reasoning Stateful Synthetic Dialogue (SSD) from visual inputs. By orchestrating iterative problem-solving, where peer student agents engage in bidirectional critique under the ground-truth-anchored guidance of a teacher, VisMind transforms static, fragmented responses into rich, multi-turn logical derivations. Utilizing this interactive paradigm, we construct a dense mid-training corpus of approximately 1.14B tokens. Extensive experiments across the Qwen2-VL and InternVL3 model families demonstrate that applying this corpus via Continued Pretraining consistently outperforms a representative stateless synthesis approach and elevates standard baselines by up to a 2.24% absolute average gain across comprehensive mathematical and logical benchmarks. Furthermore, in-depth analysis reveals that the SSD paradigm exhibits continuous linear scaling without early saturation, fosters intrinsic reasoning generalization, and establishes a robust cognitive foundation that enhances the efficacy of subsequent Reinforcement Learning with Verifiable Rewards.</abstract>
    <identifier type="citekey">fu-etal-2026-vismind</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1158/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>24493</start>
            <end>24510</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T VisMind: Stateful Synthetic Dialogue for Visual Mid-Training
%A Fu, Xinlong
%A Zhou, Zihao
%A Wang, Qiufeng
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%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 fu-etal-2026-vismind
%X The mid-training phase has emerged as a critical bridge for enhancing the reasoning capabilities of Vision-Language Models, yet its effectiveness is constrained by the scarcity of high-quality visual reasoning data. While data synthesis offers a promising alternative, conventional stateless synthesis paradigms produce shallow trajectories that quickly succumb to utility plateaus. To overcome this, we propose VisMind, a novel multi-agent collaborative framework that generates deep-reasoning Stateful Synthetic Dialogue (SSD) from visual inputs. By orchestrating iterative problem-solving, where peer student agents engage in bidirectional critique under the ground-truth-anchored guidance of a teacher, VisMind transforms static, fragmented responses into rich, multi-turn logical derivations. Utilizing this interactive paradigm, we construct a dense mid-training corpus of approximately 1.14B tokens. Extensive experiments across the Qwen2-VL and InternVL3 model families demonstrate that applying this corpus via Continued Pretraining consistently outperforms a representative stateless synthesis approach and elevates standard baselines by up to a 2.24% absolute average gain across comprehensive mathematical and logical benchmarks. Furthermore, in-depth analysis reveals that the SSD paradigm exhibits continuous linear scaling without early saturation, fosters intrinsic reasoning generalization, and establishes a robust cognitive foundation that enhances the efficacy of subsequent Reinforcement Learning with Verifiable Rewards.
%U https://aclanthology.org/2026.findings-emnlp.1158/
%P 24493-24510

Markdown (Informal)

[VisMind: Stateful Synthetic Dialogue for Visual Mid-Training](https://aclanthology.org/2026.findings-emnlp.1158/) (Fu et al., Findings 2026)

• VisMind: Stateful Synthetic Dialogue for Visual Mid-Training (Fu et al., Findings 2026)

ACL

• Xinlong Fu, Zihao Zhou, and Qiufeng Wang. 2026. VisMind: Stateful Synthetic Dialogue for Visual Mid-Training. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 24493–24510, Budapest, Hungary. Association for Computational Linguistics.