<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yari-koto-2026-amir">
    <titleInfo>
        <title>AMIR-GRPO: Inducing Implicit Preference Signals into GRPO</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Amir</namePart>
        <namePart type="given">Hossein</namePart>
        <namePart type="family">Yari</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Fajri</namePart>
        <namePart type="family">Koto</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>The Proceedings of the 4th Workshop on Mathematical Natural Language Processing (MathNLP 2026)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Marco</namePart>
            <namePart type="family">Valentino</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Deborah</namePart>
            <namePart type="family">Ferreira</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mokanarangan</namePart>
            <namePart type="family">Thayaparan</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Leonardo</namePart>
            <namePart type="family">Ranaldi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jo-Ku</namePart>
            <namePart type="family">Cheng</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Andre</namePart>
            <namePart type="family">Freitas</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">979-8-89176-468-2</identifier>
    </relatedItem>
    <abstract>Despite its effectiveness for post-training reasoning models, group relative policy optimization (GRPO) suffers from several key limitations. Specifically, it introduces response-level length bias, weakly penalizes low-quality trajectories, and reduces rich intra-group preference information to a simplified scalar reward, limiting direct pairwise comparison between reasoning trajectories. To address these issues, we propose AMIR-GRPO, which augments GRPO with an implicit DPO-style contrastive regularizer constructed directly from intra-group reward rankings, without requiring additional annotations. This mechanism strengthens suppression of low-reward trajectories, mitigates length bias, and converts each rollout group into a denser set of supervision constraints. Across multiple mathematical reasoning benchmarks, AMIR-GRPO consistently outperforms strong GRPO baselines, achieves clearer separation between correct and incorrect reasoning chains, and delivers broader coverage beyond instances solved by standard GRPO.</abstract>
    <identifier type="citekey">yari-koto-2026-amir</identifier>
    <location>
        <url>https://aclanthology.org/2026.mathnlp-main.6/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>69</start>
            <end>85</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T AMIR-GRPO: Inducing Implicit Preference Signals into GRPO
%A Yari, Amir Hossein
%A Koto, Fajri
%Y Valentino, Marco
%Y Ferreira, Deborah
%Y Thayaparan, Mokanarangan
%Y Ranaldi, Leonardo
%Y Cheng, Jo-Ku
%Y Freitas, Andre
%S The Proceedings of the 4th Workshop on Mathematical Natural Language Processing (MathNLP 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-468-2
%F yari-koto-2026-amir
%X Despite its effectiveness for post-training reasoning models, group relative policy optimization (GRPO) suffers from several key limitations. Specifically, it introduces response-level length bias, weakly penalizes low-quality trajectories, and reduces rich intra-group preference information to a simplified scalar reward, limiting direct pairwise comparison between reasoning trajectories. To address these issues, we propose AMIR-GRPO, which augments GRPO with an implicit DPO-style contrastive regularizer constructed directly from intra-group reward rankings, without requiring additional annotations. This mechanism strengthens suppression of low-reward trajectories, mitigates length bias, and converts each rollout group into a denser set of supervision constraints. Across multiple mathematical reasoning benchmarks, AMIR-GRPO consistently outperforms strong GRPO baselines, achieves clearer separation between correct and incorrect reasoning chains, and delivers broader coverage beyond instances solved by standard GRPO.
%U https://aclanthology.org/2026.mathnlp-main.6/
%P 69-85

Markdown (Informal)

[AMIR-GRPO: Inducing Implicit Preference Signals into GRPO](https://aclanthology.org/2026.mathnlp-main.6/) (Yari & Koto, MathNLP 2026)

• AMIR-GRPO: Inducing Implicit Preference Signals into GRPO (Yari & Koto, MathNLP 2026)

ACL

• Amir Hossein Yari and Fajri Koto. 2026. AMIR-GRPO: Inducing Implicit Preference Signals into GRPO. In The Proceedings of the 4th Workshop on Mathematical Natural Language Processing (MathNLP 2026), pages 69–85, Budapest, Hungary. Association for Computational Linguistics.