FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF

FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF | 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 Research Article FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF Flint Xiaofeng Fan, Cheston Tan, Yew-Soon Ong, Roger Wattenhofer, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6538432/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 In the era of increasing privacy concerns and demand for personalized experiences, traditional Reinforcement Learning with Human Feedback (RLHF) frameworks face significant challenges due to their reliance on centralized data. We introduce Federated Reinforcement Learning with Human Feedback (FedRLHF), a novel framework that decentralizes the RLHF process. FedRLHF enables collaborative policy learning across multiple clients, such as Large Language Models (LLMs) finetuning, without sharing raw data or human feedback, thereby ensuring robust privacy preservation. Leveraging federated reinforcement learning, each client integrates human feedback locally into reward functions and updates their policies through personalized RLHF processes. We establish rigorous theoretical foundations for FedRLHF, providing convergence guarantees, and deriving sample complexity bounds that scale efficiently with the number of clients. Empirical evaluations on the MovieLens and IMDb datasets demonstrate that FedRLHF preserves user privacy, achieves performance on par with centralized RLHF, and enhances personalization across diverse client environments. Artificial Intelligence and Machine Learning Federated RL RLHF LLMs Personalization Privacy-preserving AI 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-6538432","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":448536252,"identity":"f9a08120-e10e-4014-8926-b8b59df7d054","order_by":0,"name":"Flint Xiaofeng Fan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAo0lEQVRIiWNgGAWjYBAC9gbGhw9gHAmitPAcYDY2IFmLGVwlkVokktmqeRjuJM5vYD54m4coLTyH2YAqnyVuOMCWbE2UFnv2/mNALYcTNzDwmEkTZwszM1sxSMv8Bv5vRGphb2ZjBmlpOMDDRqQWnsPMknMMDhtvOMxmbDmHKC0SyYwf3lQclp3f3vzwxhtitEAAKDKZiVc+CkbBKBgFo4AQAABQ9SivGX63mQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-1658-4699","institution":"Agency for Science, Technology and Research","correspondingAuthor":true,"prefix":"","firstName":"Flint","middleName":"Xiaofeng","lastName":"Fan","suffix":""},{"id":448536362,"identity":"8944791e-742d-4f50-bee2-882a7c5128d8","order_by":1,"name":"Cheston Tan","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Cheston","middleName":"","lastName":"Tan","suffix":""},{"id":448536363,"identity":"f4409da0-57fa-428b-9aed-9ef8bc5be22f","order_by":2,"name":"Yew-Soon Ong","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Yew-Soon","middleName":"","lastName":"Ong","suffix":""},{"id":448536364,"identity":"fcf63a08-aa0e-4bae-b425-3f4206fe407f","order_by":3,"name":"Roger Wattenhofer","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Roger","middleName":"","lastName":"Wattenhofer","suffix":""},{"id":448536365,"identity":"be9651b5-ec88-44ae-a2c0-d692f013d8c7","order_by":4,"name":"Wei-Tsang Ooi","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Wei-Tsang","middleName":"","lastName":"Ooi","suffix":""}],"badges":[],"createdAt":"2025-04-27 06:38:05","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-6538432/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6538432/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81602137,"identity":"0ae968f0-24fb-4e6d-8ba1-2c11715def51","added_by":"auto","created_at":"2025-04-29 04:33:43","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":809382,"visible":true,"origin":"","legend":"","description":"","filename":"FedRLHFAAMAS2025Cameraready6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6538432/v1_covered_5b73c0c8-ec31-4d22-ad0f-9bdbf2ddfdb6.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eFedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","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":"Federated RL, RLHF, LLMs, Personalization, Privacy-preserving AI","lastPublishedDoi":"10.21203/rs.3.rs-6538432/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6538432/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn the era of increasing privacy concerns and demand for personalized experiences, traditional Reinforcement Learning with Human Feedback (RLHF) frameworks face significant challenges due to their reliance on centralized data. We introduce Federated Reinforcement Learning with Human Feedback (FedRLHF), a novel framework that decentralizes the RLHF process. FedRLHF enables collaborative policy learning across multiple clients, such as Large Language Models (LLMs) finetuning, without sharing raw data or human feedback, thereby ensuring robust privacy preservation. Leveraging federated reinforcement learning, each client integrates human feedback locally into reward functions and updates their policies through personalized RLHF processes. We establish rigorous theoretical foundations for FedRLHF, providing convergence guarantees, and deriving sample complexity bounds that scale efficiently with the number of clients. Empirical evaluations on the MovieLens and IMDb datasets demonstrate that FedRLHF preserves user privacy, achieves performance on par with centralized RLHF, and enhances personalization across diverse client environments.\u0026nbsp;\u003c/p\u003e","manuscriptTitle":"FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-29 04:25:34","doi":"10.21203/rs.3.rs-6538432/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":"784d5c65-6409-45e2-8e71-76a2ff98cb9b","owner":[],"postedDate":"April 29th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":47734378,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-04-29T04:25:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-29 04:25:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6538432","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6538432","identity":"rs-6538432","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}