<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zheng-etal-2025-asymmetric">
    <titleInfo>
        <title>Asymmetric Conflict and Synergy in Post-training for LLM-based Multilingual Machine Translation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Tong</namePart>
        <namePart type="family">Zheng</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yan</namePart>
        <namePart type="family">Wen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Huiwen</namePart>
        <namePart type="family">Bao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Junfeng</namePart>
        <namePart type="family">Guo</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Heng</namePart>
        <namePart type="family">Huang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: ACL 2025</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Wanxiang</namePart>
            <namePart type="family">Che</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Joyce</namePart>
            <namePart type="family">Nabende</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ekaterina</namePart>
            <namePart type="family">Shutova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mohammad</namePart>
            <namePart type="given">Taher</namePart>
            <namePart type="family">Pilehvar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Vienna, Austria</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-256-5</identifier>
    </relatedItem>
    <abstract>The emergence of Large Language Models (LLMs) has advanced the multilingual machine translation (MMT), yet the Curse of Multilinguality (CoM) remains a major challenge. Existing work in LLM-based MMT typically mitigates this issue via scaling up training and computation budget, which raises a critical question: Is scaling up the training and computation budget truly necessary for high-quality MMT, or can a deeper understanding of CoM provide a more efficient solution? To explore this problem, we analyze the linguistic conflicts and synergy, the underlying mechanism of CoM during post-training phase. We identify an asymmetric phenomenon in linguistic conflicts and synergy: the dominance of conflicts and synergy varies in different translation directions, leading to sub-optimal adaptation in existing post-training methods. We further find that a significant bottleneck in MMT appears to lie in post-training rather than multilingual pre-training, suggesting the need for more effective adaptation strategies. Building on these new insights, we propose a direction-aware training approach, combined with group-wise model merging, to address asymmetry in linguistic conflicts and synergy explicitly. Leveraging this strategy, our method fine-tunes X-ALMA-13B-Pretrain—trained only with multilingual pre-training—achieving comparable performance to XALMA-13B (only SFT) while using only 20B pretraining tokens and 17B parameters—5.5× fewer pretraining-tokens and 1.7x fewer model size—with just 0.85 COMET drop on Flores-200 testsets of 50 languages.</abstract>
    <identifier type="citekey">zheng-etal-2025-asymmetric</identifier>
    <identifier type="doi">10.18653/v1/2025.findings-acl.944</identifier>
    <location>
        <url>https://aclanthology.org/2025.findings-acl.944/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>18362</start>
            <end>18383</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Asymmetric Conflict and Synergy in Post-training for LLM-based Multilingual Machine Translation
%A Zheng, Tong
%A Wen, Yan
%A Bao, Huiwen
%A Guo, Junfeng
%A Huang, Heng
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F zheng-etal-2025-asymmetric
%X The emergence of Large Language Models (LLMs) has advanced the multilingual machine translation (MMT), yet the Curse of Multilinguality (CoM) remains a major challenge. Existing work in LLM-based MMT typically mitigates this issue via scaling up training and computation budget, which raises a critical question: Is scaling up the training and computation budget truly necessary for high-quality MMT, or can a deeper understanding of CoM provide a more efficient solution? To explore this problem, we analyze the linguistic conflicts and synergy, the underlying mechanism of CoM during post-training phase. We identify an asymmetric phenomenon in linguistic conflicts and synergy: the dominance of conflicts and synergy varies in different translation directions, leading to sub-optimal adaptation in existing post-training methods. We further find that a significant bottleneck in MMT appears to lie in post-training rather than multilingual pre-training, suggesting the need for more effective adaptation strategies. Building on these new insights, we propose a direction-aware training approach, combined with group-wise model merging, to address asymmetry in linguistic conflicts and synergy explicitly. Leveraging this strategy, our method fine-tunes X-ALMA-13B-Pretrain—trained only with multilingual pre-training—achieving comparable performance to XALMA-13B (only SFT) while using only 20B pretraining tokens and 17B parameters—5.5× fewer pretraining-tokens and 1.7x fewer model size—with just 0.85 COMET drop on Flores-200 testsets of 50 languages.
%R 10.18653/v1/2025.findings-acl.944
%U https://aclanthology.org/2025.findings-acl.944/
%U https://doi.org/10.18653/v1/2025.findings-acl.944
%P 18362-18383

Markdown (Informal)

[Asymmetric Conflict and Synergy in Post-training for LLM-based Multilingual Machine Translation](https://aclanthology.org/2025.findings-acl.944/) (Zheng et al., Findings 2025)

• Asymmetric Conflict and Synergy in Post-training for LLM-based Multilingual Machine Translation (Zheng et al., Findings 2025)

ACL

• Tong Zheng, Yan Wen, Huiwen Bao, Junfeng Guo, and Heng Huang. 2025. Asymmetric Conflict and Synergy in Post-training for LLM-based Multilingual Machine Translation. In Findings of the Association for Computational Linguistics: ACL 2025, pages 18362–18383, Vienna, Austria. Association for Computational Linguistics.