<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="creanga-dinu-2026-warmed">
    <titleInfo>
        <title>Who Warmed the Archives? LLMs Overestimate Historical Warmth</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Claudiu</namePart>
        <namePart type="family">Creanga</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Liviu</namePart>
        <namePart type="given">P</namePart>
        <namePart type="family">Dinu</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 3rd Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2026)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Tobias</namePart>
            <namePart type="family">Schimanski</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Aida</namePart>
            <namePart type="family">Usmanova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Gaku</namePart>
            <namePart type="family">Morio</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jingwei</namePart>
            <namePart type="family">Ni</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Kalyan</namePart>
            <namePart type="family">Dutia</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Seyed</namePart>
            <namePart type="given">Saeid</namePart>
            <namePart type="family">Vaghefi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Dominik</namePart>
            <namePart type="family">Stammbach</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Markus</namePart>
            <namePart type="family">Leippold</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">979-8-89176-465-1</identifier>
    </relatedItem>
    <abstract>Historical archives are an under-used source for extending the instrumental climate record backward in time, and LLMs offer a way to extract the indices climatologists derive by hand. Beyond measuring how well systems extract this signal, we check whether their errors are safe to use for cross-century comparison, since a good correlation score does not rule out systematic, era-linked bias. Comparing lexical baselines, fine-tuned historical transformers, and LLM prompting on the Pfister temperature index across five centuries of German text, lexical methods beat every fine-tuned transformer we test, including one pretrained from scratch on historical German (r=-0.016). All six LLMs we test (Gemini 2.5 Flash, GPT-5-mini, DeepSeek v4 Flash, Claude Sonnet 4.6, Qwen3.7-Plus, Kimi-K2.6-Fast) show a warm bias that grows with calendar year, with the same sign in every model (slopes +0.13 to +0.34/century, p&lt;0.01). The effect is modest in size (r²\approx0.01–0.05) but consistent across six independently developed models. The best-correlated of the six, Gemini 2.5 Flash, matches the best lexical correlation (r=0.32) at double the error. An ablation stripping explicit dates and calendar-era markers from the quotes leaves this trend essentially unchanged, favoring an anachronistic present-day prior over the model correctly inferring the quote’s era. Correlation alone is thus insufficient for vetting an LLM as a historical-climate-index oracle.</abstract>
    <identifier type="citekey">creanga-dinu-2026-warmed</identifier>
    <location>
        <url>https://aclanthology.org/2026.climatenlp-1.4/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>41</start>
            <end>50</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Who Warmed the Archives? LLMs Overestimate Historical Warmth
%A Creanga, Claudiu
%A Dinu, Liviu P.
%Y Schimanski, Tobias
%Y Usmanova, Aida
%Y Morio, Gaku
%Y Ni, Jingwei
%Y Dutia, Kalyan
%Y Vaghefi, Seyed Saeid
%Y Stammbach, Dominik
%Y Leippold, Markus
%S Proceedings of the 3rd Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-465-1
%F creanga-dinu-2026-warmed
%X Historical archives are an under-used source for extending the instrumental climate record backward in time, and LLMs offer a way to extract the indices climatologists derive by hand. Beyond measuring how well systems extract this signal, we check whether their errors are safe to use for cross-century comparison, since a good correlation score does not rule out systematic, era-linked bias. Comparing lexical baselines, fine-tuned historical transformers, and LLM prompting on the Pfister temperature index across five centuries of German text, lexical methods beat every fine-tuned transformer we test, including one pretrained from scratch on historical German (r=-0.016). All six LLMs we test (Gemini 2.5 Flash, GPT-5-mini, DeepSeek v4 Flash, Claude Sonnet 4.6, Qwen3.7-Plus, Kimi-K2.6-Fast) show a warm bias that grows with calendar year, with the same sign in every model (slopes +0.13 to +0.34/century, p<0.01). The effect is modest in size (r²\approx0.01–0.05) but consistent across six independently developed models. The best-correlated of the six, Gemini 2.5 Flash, matches the best lexical correlation (r=0.32) at double the error. An ablation stripping explicit dates and calendar-era markers from the quotes leaves this trend essentially unchanged, favoring an anachronistic present-day prior over the model correctly inferring the quote’s era. Correlation alone is thus insufficient for vetting an LLM as a historical-climate-index oracle.
%U https://aclanthology.org/2026.climatenlp-1.4/
%P 41-50

Markdown (Informal)

[Who Warmed the Archives? LLMs Overestimate Historical Warmth](https://aclanthology.org/2026.climatenlp-1.4/) (Creanga & Dinu, ClimateNLP 2026)

• Who Warmed the Archives? LLMs Overestimate Historical Warmth (Creanga & Dinu, ClimateNLP 2026)

ACL

• Claudiu Creanga and Liviu P Dinu. 2026. Who Warmed the Archives? LLMs Overestimate Historical Warmth. In Proceedings of the 3rd Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2026), pages 41–50, Budapest, Hungary. Association for Computational Linguistics.