<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wan-etal-2026-multilingual-chain">
    <titleInfo>
        <title>Multilingual Chain of Thought Compression via Cross-lingual Entropy Gap</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jiarui</namePart>
        <namePart type="family">Wan</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yufeng</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Songming</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rui</namePart>
        <namePart type="family">Qi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xue</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jinan</namePart>
        <namePart type="family">Xu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kaiyu</namePart>
        <namePart type="family">Huang</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>Current large reasoning models (LRMs) prefer to generate verbose chain-of-thought (CoT) traces (a.k.a. the “overthinking” issue), especially in multilingual scenarios due to the imbalanced proficiency across different languages. However, existing studies on CoT compression focus exclusively on English and overlook language characteristics, resulting in performance degradation in multilingual scenarios. To address this issue, we develop a multilingual CoT compression method based on reinforcement learning, named Language-aware Entropy-based (LAE), which optimizes LLMs with multilingual group sampling and entropy gap-based rewards. Specifically, we extend GRPO with multilingual group sampling, where each query is jointly optimized across multiple languages. We further design a language-aware entropy-gap reward that penalizes redundant uncertainty in the thinking phase and aligns reasoning complexity across languages, enabling concise yet accurate multilingual CoT reasoning. Extensive experiments on the multilingual reasoning benchmarks MGSM, MMATH and MMMLU demonstrate that our method achieves competitive performance and generalization, compared with several strong baselines.</abstract>
    <identifier type="citekey">wan-etal-2026-multilingual-chain</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2539/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>57744</start>
            <end>57764</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Multilingual Chain of Thought Compression via Cross-lingual Entropy Gap
%A Wan, Jiarui
%A Chen, Yufeng
%A Zhang, Songming
%A Qi, Rui
%A Zhang, Xue
%A Xu, Jinan
%A Huang, Kaiyu
%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 wan-etal-2026-multilingual-chain
%X Current large reasoning models (LRMs) prefer to generate verbose chain-of-thought (CoT) traces (a.k.a. the “overthinking” issue), especially in multilingual scenarios due to the imbalanced proficiency across different languages. However, existing studies on CoT compression focus exclusively on English and overlook language characteristics, resulting in performance degradation in multilingual scenarios. To address this issue, we develop a multilingual CoT compression method based on reinforcement learning, named Language-aware Entropy-based (LAE), which optimizes LLMs with multilingual group sampling and entropy gap-based rewards. Specifically, we extend GRPO with multilingual group sampling, where each query is jointly optimized across multiple languages. We further design a language-aware entropy-gap reward that penalizes redundant uncertainty in the thinking phase and aligns reasoning complexity across languages, enabling concise yet accurate multilingual CoT reasoning. Extensive experiments on the multilingual reasoning benchmarks MGSM, MMATH and MMMLU demonstrate that our method achieves competitive performance and generalization, compared with several strong baselines.
%U https://aclanthology.org/2026.emnlp-main.2539/
%P 57744-57764

Markdown (Informal)

[Multilingual Chain of Thought Compression via Cross-lingual Entropy Gap](https://aclanthology.org/2026.emnlp-main.2539/) (Wan et al., EMNLP 2026)

• Multilingual Chain of Thought Compression via Cross-lingual Entropy Gap (Wan et al., EMNLP 2026)

ACL

• Jiarui Wan, Yufeng Chen, Songming Zhang, Rui Qi, Xue Zhang, Jinan Xu, and Kaiyu Huang. 2026. Multilingual Chain of Thought Compression via Cross-lingual Entropy Gap. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 57744–57764, Budapest, Hungary. Association for Computational Linguistics.