<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sun-etal-2025-self">
    <titleInfo>
        <title>The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding?</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yutao</namePart>
        <namePart type="family">Sun</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Mingshuai</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Tiancheng</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ruochen</namePart>
        <namePart type="family">Xu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zilun</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jianwei</namePart>
        <namePart type="family">Yin</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>Self-improving large language models (LLMs) – i.e., to improve the performance of an LLM by fine-tuning it with synthetic data generated by itself – is a promising way to advance the capabilities of LLMs while avoiding extensive supervision. Existing approaches to self-improvement often rely on external supervision signals in the form of seed data and/or assistance from third-party models. This paper presents Crescent – a simple yet effective framework for generating high-quality synthetic question-answer data in a fully autonomous manner. Crescent first elicits the LLM to generate raw questions via a bait prompt, then diversifies these questions leveraging a rejection sampling-based self-deduplication, and finally feeds the questions to the LLM and collects the corresponding answers by means of majority voting. We show that Crescent sheds light on the potential of true self-improvement with zero external supervision signals for math reasoning; in particular, Crescent-generated question-answer pairs suffice to (i) improve the reasoning capabilities of an LLM while preserving its general performance (especially in the 0-shot setting); and (ii) distill LLM knowledge to weaker models more effectively than existing methods based on seed-dataset augmentation.</abstract>
    <identifier type="citekey">sun-etal-2025-self</identifier>
    <identifier type="doi">10.18653/v1/2025.findings-acl.337</identifier>
    <location>
        <url>https://aclanthology.org/2025.findings-acl.337/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>6501</start>
            <end>6512</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding?
%A Sun, Yutao
%A Chen, Mingshuai
%A Zhao, Tiancheng
%A Xu, Ruochen
%A Zhang, Zilun
%A Yin, Jianwei
%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 sun-etal-2025-self
%X Self-improving large language models (LLMs) – i.e., to improve the performance of an LLM by fine-tuning it with synthetic data generated by itself – is a promising way to advance the capabilities of LLMs while avoiding extensive supervision. Existing approaches to self-improvement often rely on external supervision signals in the form of seed data and/or assistance from third-party models. This paper presents Crescent – a simple yet effective framework for generating high-quality synthetic question-answer data in a fully autonomous manner. Crescent first elicits the LLM to generate raw questions via a bait prompt, then diversifies these questions leveraging a rejection sampling-based self-deduplication, and finally feeds the questions to the LLM and collects the corresponding answers by means of majority voting. We show that Crescent sheds light on the potential of true self-improvement with zero external supervision signals for math reasoning; in particular, Crescent-generated question-answer pairs suffice to (i) improve the reasoning capabilities of an LLM while preserving its general performance (especially in the 0-shot setting); and (ii) distill LLM knowledge to weaker models more effectively than existing methods based on seed-dataset augmentation.
%R 10.18653/v1/2025.findings-acl.337
%U https://aclanthology.org/2025.findings-acl.337/
%U https://doi.org/10.18653/v1/2025.findings-acl.337
%P 6501-6512

Markdown (Informal)

[The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding?](https://aclanthology.org/2025.findings-acl.337/) (Sun et al., Findings 2025)

• The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding? (Sun et al., Findings 2025)

ACL

• Yutao Sun, Mingshuai Chen, Tiancheng Zhao, Ruochen Xu, Zilun Zhang, and Jianwei Yin. 2025. The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding?. In Findings of the Association for Computational Linguistics: ACL 2025, pages 6501–6512, Vienna, Austria. Association for Computational Linguistics.