Position: Static Benchmarks and Accuracy-Centric Metrics Are Inadequate for Responsible Foundation Models: A Call for Holistic Evaluation Integrating Risk Factors, Performance, and Governance Frameworks

Position: Static Benchmarks and Accuracy-Centric Metrics Are Inadequate for Responsible Foundation Models: A Call for Holistic Evaluation Integrating Risk Factors, Performance, and Governance Frameworks | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Short Report Position: Static Benchmarks and Accuracy-Centric Metrics Are Inadequate for Responsible Foundation Models: A Call for Holistic Evaluation Integrating Risk Factors, Performance, and Governance Frameworks Rizwan Qureshi, Shaina Raza, Anas Zafar, Amgad Muneer, Sumair Soomro, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8927316/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This position paper argues that reliance on static benchmarks and accuracy-centric metrics is insufficient for assessing the responsibility of foundation models. High benchmark performance is often treated as a proxy for trustworthiness, yet accuracy alone provides limited insight into critical risks such as bias, privacy leakage, unsafe behavior, loss of trust, and governance non-compliance. We propose a holistic evaluation perspective that jointly considers task performance, risk factors, and governance alignment. This view is formalized through a composite Responsible AI (RAI) formulation and illustrates how models with similar accuracy can differ substantially in real-world acceptability. By examining limitations of current evaluation practices and engaging alternative viewpoints, this paper encourages evaluation approaches that better reflect deployment realities. Artificial Intelligence and Machine Learning Foundation models responsible AI evaluation benchmarks risk governance Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8927316","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Short Report","associatedPublications":[],"authors":[{"id":594539442,"identity":"c1ca2960-8805-471a-b3eb-ea83116dc308","order_by":0,"name":"Rizwan Qureshi","email":"","orcid":"","institution":"Salim Habib University","correspondingAuthor":false,"prefix":"","firstName":"Rizwan","middleName":"","lastName":"Qureshi","suffix":""},{"id":594539443,"identity":"1bcfde79-e66e-4561-8e88-065ab5318136","order_by":1,"name":"Shaina Raza","email":"","orcid":"","institution":"Vector Institute","correspondingAuthor":false,"prefix":"","firstName":"Shaina","middleName":"","lastName":"Raza","suffix":""},{"id":594539444,"identity":"db97af1c-80f6-4ea7-afc7-2fb33efb60fd","order_by":2,"name":"Anas Zafar","email":"","orcid":"","institution":"The University of Texas MD Anderson Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Anas","middleName":"","lastName":"Zafar","suffix":""},{"id":594539445,"identity":"e2dcb339-89d5-47b7-8046-9e9909820f3b","order_by":3,"name":"Amgad Muneer","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYJADwwdAgoePFC3GBiAtbKRoMZMAkQS16M7IPfjg4x67aHn3w9sqv+bYybAxMD98dAOfyTfykg1nPEvO3Xgmrey27LZkoMPYjI1z8GrJMZPmOcCcu7Ehx+y25DZmoBYeNmkCWsx//zlQn7ux/41ZseS2eqK0mDEzHDicO18ix4zx47bDRGg588ZYsufA8dwNEs+KpRm3HedhYybkl+M5hh9+HKjOnd+fvPHjz23V9vzszQ8f49MCBwYHGBiYeUAsZmKUg4B8AwMD4w9iVY+CUTAKRsGIAgC3KUpn3vXL5AAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-7157-3020","institution":"The University of Texas MD Anderson Cancer Center","correspondingAuthor":true,"prefix":"","firstName":"Amgad","middleName":"","lastName":"Muneer","suffix":""},{"id":594539446,"identity":"8d874898-1716-47c7-a9ce-73e1e85fae49","order_by":4,"name":"Sumair Soomro","email":"","orcid":"","institution":"University of Massachusetts","correspondingAuthor":false,"prefix":"","firstName":"Sumair","middleName":"","lastName":"Soomro","suffix":""},{"id":594539447,"identity":"b32cfd5e-6ab3-4021-8671-874b0e8a2dc3","order_by":5,"name":"Umaima Khan","email":"","orcid":"","institution":"University of Central Florida","correspondingAuthor":false,"prefix":"","firstName":"Umaima","middleName":"","lastName":"Khan","suffix":""},{"id":594539448,"identity":"4b933935-30d5-4097-bffc-8c33961d030f","order_by":6,"name":"Ferhat Sadak","email":"","orcid":"","institution":"Aston University","correspondingAuthor":false,"prefix":"","firstName":"Ferhat","middleName":"","lastName":"Sadak","suffix":""}],"badges":[],"createdAt":"2026-02-20 15:27:46","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8927316/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8927316/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104402518,"identity":"2a3bbf68-ebaa-40a6-a18b-f1e3e7bda173","added_by":"auto","created_at":"2026-03-11 12:15:36","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1815249,"visible":true,"origin":"","legend":"","description":"","filename":"PositionStaticBenchmarksandAccuracyCentricMetricsAreInadequateforResponsibleFoundationModelsACallforHolisticEvaluationIntegrating.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8927316/v1_covered_5057c023-a2dd-41b0-aedd-4977133e36db.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003ePosition: Static Benchmarks and Accuracy-Centric Metrics Are Inadequate for Responsible Foundation Models: A Call for Holistic Evaluation Integrating Risk Factors, Performance, and Governance Frameworks\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Vector Institute","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Foundation models, responsible AI, evaluation, benchmarks, risk, governance","lastPublishedDoi":"10.21203/rs.3.rs-8927316/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8927316/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis position paper argues that reliance on static benchmarks and accuracy-centric metrics is insufficient for assessing the responsibility of foundation models. High benchmark performance is often treated as a proxy for trustworthiness, yet accuracy alone provides limited insight into critical risks such as bias, privacy leakage, unsafe behavior, loss of trust, and governance non-compliance. We propose a holistic evaluation perspective that jointly considers task performance, risk factors, and governance alignment. This view is formalized through a composite Responsible AI (RAI) formulation and illustrates how models with similar accuracy can differ substantially in real-world acceptability. By examining limitations of current evaluation practices and engaging alternative viewpoints, this paper encourages evaluation approaches that better reflect deployment realities.\u003c/p\u003e","manuscriptTitle":"Position: Static Benchmarks and Accuracy-Centric Metrics Are Inadequate for Responsible Foundation Models: A Call for Holistic Evaluation Integrating Risk Factors, Performance, and Governance Frameworks","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-05 07:13:43","doi":"10.21203/rs.3.rs-8927316/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"db1ab3ee-29f3-4f75-a56e-44e21be49692","owner":[],"postedDate":"March 5th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63271674,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-03-05T07:13:44+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-05 07:13:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8927316","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8927316","identity":"rs-8927316","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}