<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dai-etal-2026-goal">
    <titleInfo>
        <title>M-GOAL: Toward Accurate and Language-Consistent Reasoning in LLMs via English-Only Post-Training</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Dingyuan</namePart>
        <namePart type="family">Dai</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yuxi</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiayuan</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>Reasoning-oriented language models inherit strong English-centric biases from pre-training, which are further reinforced by English-only post-training. This creates a dual failure mode where models fail both to reason correctly and to maintain reasoning traces in the query language. We study reasoning accuracy and language consistency under English-only post-training. SFT primarily improves language consistency with limited or inconsistent accuracy gains, while group relative policy optimization (GRPO) substantially improves accuracy at the cost of language consistency. Although translated multilingual GRPO improves both objectives, it requires translating reasoning post-training data into multiple languages and may introduce translation noise and information loss. We find that the language-consistency capability acquired by SFT and the accuracy capability acquired by GRPO are approximately orthogonal in the model’s update space. By merging an English LoRA-SFT adapter with an English GRPO checkpoint trained from the same base model, M-GOAL improves MGSM accuracy from 39.1 to 57.8 and language consistency from 17.1 to 75.2 without translated reasoning supervision. This composition outperforms sequential training, its merge coefficient exposes a tunable operating point along the accuracy–consistency frontier, and the pattern holds beyond mathematical reasoning and on benchmarks whose non-English problems are natively authored rather than translated. Our results suggest that multilingual reasoning post-training does not necessarily require multilingual post-training data.</abstract>
    <identifier type="citekey">dai-etal-2026-goal</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.1232</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1232/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>26066</start>
            <end>26084</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T M-GOAL: Toward Accurate and Language-Consistent Reasoning in LLMs via English-Only Post-Training
%A Dai, Dingyuan
%A Chen, Yuxi
%A Wang, Jiayuan
%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 dai-etal-2026-goal
%X Reasoning-oriented language models inherit strong English-centric biases from pre-training, which are further reinforced by English-only post-training. This creates a dual failure mode where models fail both to reason correctly and to maintain reasoning traces in the query language. We study reasoning accuracy and language consistency under English-only post-training. SFT primarily improves language consistency with limited or inconsistent accuracy gains, while group relative policy optimization (GRPO) substantially improves accuracy at the cost of language consistency. Although translated multilingual GRPO improves both objectives, it requires translating reasoning post-training data into multiple languages and may introduce translation noise and information loss. We find that the language-consistency capability acquired by SFT and the accuracy capability acquired by GRPO are approximately orthogonal in the model’s update space. By merging an English LoRA-SFT adapter with an English GRPO checkpoint trained from the same base model, M-GOAL improves MGSM accuracy from 39.1 to 57.8 and language consistency from 17.1 to 75.2 without translated reasoning supervision. This composition outperforms sequential training, its merge coefficient exposes a tunable operating point along the accuracy–consistency frontier, and the pattern holds beyond mathematical reasoning and on benchmarks whose non-English problems are natively authored rather than translated. Our results suggest that multilingual reasoning post-training does not necessarily require multilingual post-training data.
%R 10.18653/2026.findings-emnlp.1232
%U https://aclanthology.org/2026.findings-emnlp.1232/
%U https://doi.org/10.18653/2026.findings-emnlp.1232
%P 26066-26084

Markdown (Informal)

[M-GOAL: Toward Accurate and Language-Consistent Reasoning in LLMs via English-Only Post-Training](https://aclanthology.org/2026.findings-emnlp.1232/) (Dai et al., Findings 2026)

• M-GOAL: Toward Accurate and Language-Consistent Reasoning in LLMs via English-Only Post-Training (Dai et al., Findings 2026)

ACL

• Dingyuan Dai, Yuxi Chen, and Jiayuan Wang. 2026. M-GOAL: Toward Accurate and Language-Consistent Reasoning in LLMs via English-Only Post-Training. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 26066–26084, Budapest, Hungary. Association for Computational Linguistics.