<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hu-etal-2026-sa">
    <titleInfo>
        <title>SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Xue</namePart>
        <namePart type="family">Hu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Pan</namePart>
        <namePart type="family">Zewei</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zeli</namePart>
        <namePart type="family">Su</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhou</namePart>
        <namePart type="family">Liu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wentao</namePart>
        <namePart type="family">Zhang</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>LLM agents can generate paper reproduction code, yet often produce scientifically unfaithful implementations. We define this failure mode as semantic drift, where generated code silently diverges from the paper’s specifications. We introduce SemanticAlign-Bench(SA-Bench), a diagnostic benchmark covering 30 papers from ICLR, ICML and NeurIPS 2025. For each paper, we decompose its specifications into atomic and verifiable implementation claims, which we call Semantic Alignment Units (SAUs) and evaluate repositories along four diagnostic dimensions spanning numerical, methodological, protocol and ordering drift. In total, we construct 1,491 SAUs across five ML domains and evaluate 12 generator configurations (4 models 3 scaffolds). Even the strongest configuration (Claude + PaperCoder) achieves a mean SAU score of only 0.301 out of 1.0, with an overall mean of 0.221 across 360 evaluations. A failure taxonomy reveals that agents attempt most requirements but implement them incorrectly, with implementation mismatch and stubs accounting for the majority of zero-scored claims. Our analysis further indicates that scaffolds optimized for executability provide limited leverage for scientific reproduction; narrowing the gap requires scaffolds that prioritize semantic specification verification. The benchmark, annotations and evaluation pipeline are publicly available.\par Code:https://anonymous.4open.science/r/SemanticAlign-Bench-0676\par Data:https://anonymous-hf.up.railway.app/a/rrgn430zpfui/</abstract>
    <identifier type="citekey">hu-etal-2026-sa</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.170/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>3578</start>
            <end>3607</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction
%A Hu, Xue
%A Zewei, Pan
%A Su, Zeli
%A Liu, Zhou
%A Zhang, Wentao
%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 hu-etal-2026-sa
%X LLM agents can generate paper reproduction code, yet often produce scientifically unfaithful implementations. We define this failure mode as semantic drift, where generated code silently diverges from the paper’s specifications. We introduce SemanticAlign-Bench(SA-Bench), a diagnostic benchmark covering 30 papers from ICLR, ICML and NeurIPS 2025. For each paper, we decompose its specifications into atomic and verifiable implementation claims, which we call Semantic Alignment Units (SAUs) and evaluate repositories along four diagnostic dimensions spanning numerical, methodological, protocol and ordering drift. In total, we construct 1,491 SAUs across five ML domains and evaluate 12 generator configurations (4 models 3 scaffolds). Even the strongest configuration (Claude + PaperCoder) achieves a mean SAU score of only 0.301 out of 1.0, with an overall mean of 0.221 across 360 evaluations. A failure taxonomy reveals that agents attempt most requirements but implement them incorrectly, with implementation mismatch and stubs accounting for the majority of zero-scored claims. Our analysis further indicates that scaffolds optimized for executability provide limited leverage for scientific reproduction; narrowing the gap requires scaffolds that prioritize semantic specification verification. The benchmark, annotations and evaluation pipeline are publicly available.\par Code:https://anonymous.4open.science/r/SemanticAlign-Bench-0676\par Data:https://anonymous-hf.up.railway.app/a/rrgn430zpfui/
%U https://aclanthology.org/2026.findings-emnlp.170/
%P 3578-3607

Markdown (Informal)

[SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction](https://aclanthology.org/2026.findings-emnlp.170/) (Hu et al., Findings 2026)

• SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction (Hu et al., Findings 2026)

ACL

• Xue Hu, Pan Zewei, Zeli Su, Zhou Liu, and Wentao Zhang. 2026. SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 3578–3607, Budapest, Hungary. Association for Computational Linguistics.