<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="singh-2026-architecture">
    <titleInfo>
        <title>The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Prabhjot</namePart>
        <namePart type="family">Singh</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>Proceedings of the Seventh Workshop on Insights from Negative Results in NLP</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">João</namePart>
            <namePart type="family">Sedoc</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Raphael</namePart>
            <namePart type="family">Shu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Aleksandr</namePart>
            <namePart type="family">Drozd</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Arjun</namePart>
            <namePart type="family">Akula</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Anna</namePart>
            <namePart type="family">Rogers</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Salvatore</namePart>
            <namePart type="family">Giorgi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Adrian</namePart>
            <namePart type="family">Benton</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">979-8-89176-480-4</identifier>
    </relatedItem>
    <abstract>NLI debiasing research evaluates interventions on single architectures and datasets, leaving a critical assumption untested: do improvements reflect method quality or model capacity? We run 44 controlled experiments: eleven debiasing strategies across two architecturally distinct models (ELECTRA-small, RoBERTa-base) and two corpora (SNLI, MNLI), under identical conditions with Wilson 95% confidence intervals throughout. On ELECTRA-small, which achieves 87.4% in-domain accuracy, debiasing fails completely: all ten debiasing interventions fall within the baseline confidence interval on HANS (CoV = 0.75%). On RoBERTa-base, methods show genuine sensitivity (CoV = 10.76%, approximately 14x higher), yet the median intervention trends toward degrading OOD accuracy on SNLI, short of conventional significance (Wilcoxon p = 0.065), and method rankings across corpora are anti-correlated for ELECTRA (Spearman \ensuremathρ = -0.42) but stable for RoBERTa (\ensuremathρ = 0.75). Contrastive augmentation swings 5.6 pp across corpora (+3.59 pp MNLI, -2.02 pp SNLI). The positive robustness-accuracy correlation on RoBERTa (r = 0.848) inverts the assumed tradeoff: methods that improve HANS also preserve in-domain accuracy, while aggressive debiasers collapse both. The ELECTRA failure is not a pipeline artefact; the same codebase produces real gains on RoBERTa. It is a capacity failure: debiasing presupposes representational capacity that small architectures lack, a precondition the literature neither states nor tests.</abstract>
    <identifier type="citekey">singh-2026-architecture</identifier>
    <location>
        <url>https://aclanthology.org/2026.insights-1.10/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>84</start>
            <end>89</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise
%A Singh, Prabhjot
%Y Sedoc, João
%Y Shu, Raphael
%Y Drozd, Aleksandr
%Y Akula, Arjun
%Y Rogers, Anna
%Y Giorgi, Salvatore
%Y Benton, Adrian
%S Proceedings of the Seventh Workshop on Insights from Negative Results in NLP
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-480-4
%F singh-2026-architecture
%X NLI debiasing research evaluates interventions on single architectures and datasets, leaving a critical assumption untested: do improvements reflect method quality or model capacity? We run 44 controlled experiments: eleven debiasing strategies across two architecturally distinct models (ELECTRA-small, RoBERTa-base) and two corpora (SNLI, MNLI), under identical conditions with Wilson 95% confidence intervals throughout. On ELECTRA-small, which achieves 87.4% in-domain accuracy, debiasing fails completely: all ten debiasing interventions fall within the baseline confidence interval on HANS (CoV = 0.75%). On RoBERTa-base, methods show genuine sensitivity (CoV = 10.76%, approximately 14x higher), yet the median intervention trends toward degrading OOD accuracy on SNLI, short of conventional significance (Wilcoxon p = 0.065), and method rankings across corpora are anti-correlated for ELECTRA (Spearman \ensuremathρ = -0.42) but stable for RoBERTa (\ensuremathρ = 0.75). Contrastive augmentation swings 5.6 pp across corpora (+3.59 pp MNLI, -2.02 pp SNLI). The positive robustness-accuracy correlation on RoBERTa (r = 0.848) inverts the assumed tradeoff: methods that improve HANS also preserve in-domain accuracy, while aggressive debiasers collapse both. The ELECTRA failure is not a pipeline artefact; the same codebase produces real gains on RoBERTa. It is a capacity failure: debiasing presupposes representational capacity that small architectures lack, a precondition the literature neither states nor tests.
%U https://aclanthology.org/2026.insights-1.10/
%P 84-89

Markdown (Informal)

[The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise](https://aclanthology.org/2026.insights-1.10/) (Singh, insights 2026)

• The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise (Singh, insights 2026)

ACL

• Prabhjot Singh. 2026. The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise. In Proceedings of the Seventh Workshop on Insights from Negative Results in NLP, pages 84–89, Budapest, Hungary. Association for Computational Linguistics.