<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jenkins-2026-beyond">
    <titleInfo>
        <title>Beyond Static Benchmarks: A Validity, Reliability, and Sociotechnical Framework for Evaluating LLMs in Deployment Contexts</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Ben</namePart>
        <namePart type="family">Jenkins</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Workshop on Evaluating Evaluations (EvalEval)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Mubashara</namePart>
            <namePart type="family">Akhtar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jan</namePart>
            <namePart type="family">Batzner</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Leshem</namePart>
            <namePart type="family">Choshen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Avijit</namePart>
            <namePart type="family">Ghosh</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Usman</namePart>
            <namePart type="family">Gohar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jennifer</namePart>
            <namePart type="family">Mickel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ichhya</namePart>
            <namePart type="family">Pant</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Zeerak</namePart>
            <namePart type="family">Talat</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Michelle</namePart>
            <namePart type="family">Lin</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, CA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-429-3</identifier>
    </relatedItem>
    <abstract>Static leaderboards summarize large language model (LLM) performance but offer weak evidence under shifting usage, noisy inputs, and plural stakeholder values. We present VRS-Eval, operationalizing deployment validity (benchmark vs. deployment score alignment), operational reliability (stability under a declared perturbation family), and sociotechnical alignment (metric vs. elicited rubric weights as a thin audit summary). With a reproducible simulator under explicit PB vs. PD shift and multi-turn interaction, we stress-test evaluation protocols in a controlled environment: under our main setting, benchmark-side scores (on PB) exceed estimated deploymentside utility scores (evaluated on trajectories from PD) by roughly 21–26% in relative terms across three metrics, with tight 95% percentile intervals (K=200). Failure mixtures emphasize overfitting, shift fragility, and rubric misalignment, consistent with firstvs. third-party reporting asymmetries (Reuel et al., 2025). A staged pipeline narrows the validity gap and raises reliability for the same generative story. Sensitivity sweeps over |\ensuremathØmega| and rubric-label rate preserve the rank ordering of harnesses, suggesting the qualitative conclusions are robust to plausible design-choice variation within the simulator. We discuss harness and accountability implications.</abstract>
    <identifier type="citekey">jenkins-2026-beyond</identifier>
    <location>
        <url>https://aclanthology.org/2026.evaleval-1.30/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>201</start>
            <end>210</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Beyond Static Benchmarks: A Validity, Reliability, and Sociotechnical Framework for Evaluating LLMs in Deployment Contexts
%A Jenkins, Ben
%Y Akhtar, Mubashara
%Y Batzner, Jan
%Y Choshen, Leshem
%Y Ghosh, Avijit
%Y Gohar, Usman
%Y Mickel, Jennifer
%Y Pant, Ichhya
%Y Talat, Zeerak
%Y Lin, Michelle
%S Proceedings of the Workshop on Evaluating Evaluations (EvalEval)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, CA
%@ 979-8-89176-429-3
%F jenkins-2026-beyond
%X Static leaderboards summarize large language model (LLM) performance but offer weak evidence under shifting usage, noisy inputs, and plural stakeholder values. We present VRS-Eval, operationalizing deployment validity (benchmark vs. deployment score alignment), operational reliability (stability under a declared perturbation family), and sociotechnical alignment (metric vs. elicited rubric weights as a thin audit summary). With a reproducible simulator under explicit PB vs. PD shift and multi-turn interaction, we stress-test evaluation protocols in a controlled environment: under our main setting, benchmark-side scores (on PB) exceed estimated deploymentside utility scores (evaluated on trajectories from PD) by roughly 21–26% in relative terms across three metrics, with tight 95% percentile intervals (K=200). Failure mixtures emphasize overfitting, shift fragility, and rubric misalignment, consistent with firstvs. third-party reporting asymmetries (Reuel et al., 2025). A staged pipeline narrows the validity gap and raises reliability for the same generative story. Sensitivity sweeps over |\ensuremathØmega| and rubric-label rate preserve the rank ordering of harnesses, suggesting the qualitative conclusions are robust to plausible design-choice variation within the simulator. We discuss harness and accountability implications.
%U https://aclanthology.org/2026.evaleval-1.30/
%P 201-210

Markdown (Informal)

[Beyond Static Benchmarks: A Validity, Reliability, and Sociotechnical Framework for Evaluating LLMs in Deployment Contexts](https://aclanthology.org/2026.evaleval-1.30/) (Jenkins, EvalEval 2026)

• Beyond Static Benchmarks: A Validity, Reliability, and Sociotechnical Framework for Evaluating LLMs in Deployment Contexts (Jenkins, EvalEval 2026)

ACL

• Ben Jenkins. 2026. Beyond Static Benchmarks: A Validity, Reliability, and Sociotechnical Framework for Evaluating LLMs in Deployment Contexts. In Proceedings of the Workshop on Evaluating Evaluations (EvalEval), pages 201–210, San Diego, CA. Association for Computational Linguistics.