<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sharma-etal-2025-dimsum">
    <titleInfo>
        <title>DIMSUM: Discourse in Mathematical Reasoning as a Supervision Module</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Krish</namePart>
        <namePart type="family">Sharma</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Niyar</namePart>
        <namePart type="given">R</namePart>
        <namePart type="family">Barman</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Akshay</namePart>
        <namePart type="family">Chaturvedi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nicholas</namePart>
        <namePart type="family">Asher</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-08</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 26th Annual Meeting of the Special Interest Group on Discourse and Dialogue</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Frédéric</namePart>
            <namePart type="family">Béchet</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Fabrice</namePart>
            <namePart type="family">Lefèvre</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Nicholas</namePart>
            <namePart type="family">Asher</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Seokhwan</namePart>
            <namePart type="family">Kim</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Teva</namePart>
            <namePart type="family">Merlin</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Avignon, France</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>We look at reasoning on GSM8k, a dataset of short texts presenting primary school, math problems. We find, with Mirzadeh et al (2024), that current LLM progress on the data set may not be explained by better reasoning but by exposure to a broader pretraining data distribution. We then introduce a novel information source for helping models with less data or inferior training reason better: discourse structure. We show that discourse structure improves performance for models like Llama2 13b by up to 160%. Even for models that have most likely memorized the data set, adding discourse structural information to the model still improves predictions and dramatically improves large model performance on out of distribution examples.</abstract>
    <identifier type="citekey">sharma-etal-2025-dimsum</identifier>
    <location>
        <url>https://aclanthology.org/2025.sigdial-1.24/</url>
    </location>
    <part>
        <date>2025-08</date>
        <extent unit="page">
            <start>294</start>
            <end>310</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T DIMSUM: Discourse in Mathematical Reasoning as a Supervision Module
%A Sharma, Krish
%A Barman, Niyar R.
%A Chaturvedi, Akshay
%A Asher, Nicholas
%Y Béchet, Frédéric
%Y Lefèvre, Fabrice
%Y Asher, Nicholas
%Y Kim, Seokhwan
%Y Merlin, Teva
%S Proceedings of the 26th Annual Meeting of the Special Interest Group on Discourse and Dialogue
%D 2025
%8 August
%I Association for Computational Linguistics
%C Avignon, France
%F sharma-etal-2025-dimsum
%X We look at reasoning on GSM8k, a dataset of short texts presenting primary school, math problems. We find, with Mirzadeh et al (2024), that current LLM progress on the data set may not be explained by better reasoning but by exposure to a broader pretraining data distribution. We then introduce a novel information source for helping models with less data or inferior training reason better: discourse structure. We show that discourse structure improves performance for models like Llama2 13b by up to 160%. Even for models that have most likely memorized the data set, adding discourse structural information to the model still improves predictions and dramatically improves large model performance on out of distribution examples.
%U https://aclanthology.org/2025.sigdial-1.24/
%P 294-310

Markdown (Informal)

[DIMSUM: Discourse in Mathematical Reasoning as a Supervision Module](https://aclanthology.org/2025.sigdial-1.24/) (Sharma et al., SIGDIAL 2025)

• DIMSUM: Discourse in Mathematical Reasoning as a Supervision Module (Sharma et al., SIGDIAL 2025)

ACL

• Krish Sharma, Niyar R. Barman, Akshay Chaturvedi, and Nicholas Asher. 2025. DIMSUM: Discourse in Mathematical Reasoning as a Supervision Module. In Proceedings of the 26th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 294–310, Avignon, France. Association for Computational Linguistics.