ProtoMFL: A Robust Multimodal Federated Learning Framework via Cross-Modal Prototype Integration

ProtoMFL: A Robust Multimodal Federated Learning Framework via Cross-Modal Prototype Integration | 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 ProtoMFL: A Robust Multimodal Federated Learning Framework via Cross-Modal Prototype Integration Junsun Zhang, Chaochao Sun, Yuan Peng This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8376153/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Multimodal federated learning (MFL) has made substantial progress in aggregating multimodal knowledge across distributed environments. However, it still encounters persistent challenges caused by modality-missing data at the client level. Traditional knowledge distillation–based approaches provide limited performance in handling these modality-missing scenarios. To mitigate the performance degradation caused by modality dropout, this paper proposes a prototype-based multimodal federated learning framework, termed Prototype-based Multimodal Federated Learning (ProtoMFL). By replacing sample-level representations with category-level prototypes as knowledge carriers, ProtoMFL enables more efficient cross-modal knowledge aggregation. The ProtoMFL framework consists of three core components. Cross-Modal Prototype Regularisation reduces distributional discrepancies between client and global models. Cross-Modal Prototype Contrast enhances the aggregation of similar prototypes and separation of dissimilar ones through contrastive learning. Cross-Modal Alignment enforces semantic alignment between modalities at the feature level, thereby mitigating the adverse effects of modality dropout. Experimental results show that ProtoMFL significantly outperforms existing methods in both accuracy and robustness across multiple benchmark datasets. Even under severe modality dropout, ProtoMFL maintains stable performance, achieving an average improvement of approximately 2.8% over the baseline CreamFL model without prototype mechanisms. This improvement effectively mitigates model drift issues caused by heterogeneous modalities. Multimodal federated learning modality missing Prototype Learning semantic alignment Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 11 Jan, 2026 Reviews received at journal 08 Jan, 2026 Reviewers agreed at journal 27 Dec, 2025 Reviews received at journal 25 Dec, 2025 Reviewers agreed at journal 21 Dec, 2025 Reviewers agreed at journal 19 Dec, 2025 Reviewers invited by journal 17 Dec, 2025 Editor assigned by journal 16 Dec, 2025 Submission checks completed at journal 16 Dec, 2025 First submitted to journal 16 Dec, 2025 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-8376153","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":563648274,"identity":"e8d378c3-f34f-4c67-bc11-69e315f277bb","order_by":0,"name":"Junsun Zhang","email":"","orcid":"","institution":"Shanghai University of Electric Power","correspondingAuthor":false,"prefix":"","firstName":"Junsun","middleName":"","lastName":"Zhang","suffix":""},{"id":563648277,"identity":"da897f19-e279-4d29-a5b8-c1facf8a4874","order_by":1,"name":"Chaochao Sun","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAn0lEQVRIiWNgGAWjYBADOTb29gOkaTHm4zmTQJqWxHkSDgbEKTVn7z3A+KPGLr1NgiGB4UfFNsJaLHvOJTDzHEvObZNuPMDYc+Y2YS0GN3LMfzM2HMhtkzmQwMzYRpwWA8afDQfS2SQSDIjXwsDbcCCBeC0wvxi2AQP5IFF+gYWYvHx7+8EHPyqIcRgDD4JzgLB6dC2jYBSMglEwCrACAGptN90Xea5hAAAAAElFTkSuQmCC","orcid":"","institution":"Shanghai University of Electric Power","correspondingAuthor":true,"prefix":"","firstName":"Chaochao","middleName":"","lastName":"Sun","suffix":""},{"id":563648279,"identity":"f1375e63-9f00-4a3b-8c0b-c3e1d549e2c7","order_by":2,"name":"Yuan Peng","email":"","orcid":"","institution":"Shanghai University of Electric Power","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Peng","suffix":""}],"badges":[],"createdAt":"2025-12-16 12:38:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8376153/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8376153/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98730269,"identity":"5a80f7e9-90ae-4a06-b9d1-3d90e6c8237b","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5764,"visible":true,"origin":"","legend":"","description":"","filename":"65c15fb96b234a90b3bacc932f28599d.json","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/774930e2d9f075568985fa53.json"},{"id":98730270,"identity":"37997b3e-5fb2-45ef-b9a9-ab6a70321be5","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124060,"visible":true,"origin":"","legend":"","description":"","filename":"65c15fb96b234a90b3bacc932f28599d1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/22ab15bdfacf72422f46d1b3.xml"},{"id":98730272,"identity":"5e70add3-43b2-466a-96b4-6a2a3fb984d1","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1600646,"visible":true,"origin":"","legend":"","description":"","filename":"ProtoMFL.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/bf3e43b0b5c8cee311eea426.pdf"},{"id":98776491,"identity":"1e2987d6-5ece-46ec-84f8-019dc38bb8ca","added_by":"auto","created_at":"2025-12-22 12:22:58","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1660101,"visible":true,"origin":"","legend":"","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/d939255fa1f72deecf5c4428.png"},{"id":98777200,"identity":"9046149b-9598-4615-971c-bd2becf0573b","added_by":"auto","created_at":"2025-12-22 12:25:52","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2122823,"visible":true,"origin":"","legend":"","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/139322b006694ad216543fce.png"},{"id":98776594,"identity":"c6126fab-4976-47f5-bb12-79ca6f8ebc14","added_by":"auto","created_at":"2025-12-22 12:23:10","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":207255,"visible":true,"origin":"","legend":"","description":"","filename":"fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/00f6b18d86bd6b1b7db5695c.png"},{"id":98776733,"identity":"8b300d2e-ac11-45e1-a878-bab8283ac01d","added_by":"auto","created_at":"2025-12-22 12:23:25","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":245720,"visible":true,"origin":"","legend":"","description":"","filename":"fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/ae8a9f80e4e218f81fc6aa0c.png"},{"id":98730277,"identity":"3b2a942f-20cc-4281-b35e-56bc40540471","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":442313,"visible":true,"origin":"","legend":"","description":"","filename":"fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/953fc2bea2c46ecc218d8b25.png"},{"id":98777365,"identity":"2275767a-1e78-47de-9a4e-20dc6238faf2","added_by":"auto","created_at":"2025-12-22 12:26:36","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":240180,"visible":true,"origin":"","legend":"","description":"","filename":"fig6.png","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/cfb78d273453be04475ae2ce.png"},{"id":98730278,"identity":"ad125168-ae14-465b-875b-dc0e521e1414","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":421391,"visible":true,"origin":"","legend":"","description":"","filename":"snarticle.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/12aa4f697ab3aaceb9c8024f.pdf"},{"id":98778417,"identity":"1a66c3ea-697b-4d12-8b12-b3586fabc3e5","added_by":"auto","created_at":"2025-12-22 12:29:14","extension":"bst","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35515,"visible":true,"origin":"","legend":"","description":"","filename":"snbasic.bst","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/b6ca44acf2c1d1f234812a91.bst"},{"id":98730279,"identity":"34ecdafe-6008-46be-9ca5-fb63918f2b3d","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"cls","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55857,"visible":true,"origin":"","legend":"","description":"","filename":"snjnl.cls","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/5cdba9cc9f0ab638b3e419ef.cls"},{"id":98730283,"identity":"20d996bf-751a-4ab4-943c-80e4949dd6f7","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"bst","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64023,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysay.bst","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/81f9092f4be683c556caeb66.bst"},{"id":98776754,"identity":"907ce9c8-2320-4847-baf9-c2dd5568b4e2","added_by":"auto","created_at":"2025-12-22 12:23:26","extension":"bst","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64166,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysnum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/33a8bb337df1e65d78e26e73.bst"},{"id":98730276,"identity":"36b28074-d346-44e5-9116-135f4cbe0782","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":418495,"visible":true,"origin":"","legend":"","description":"","filename":"usermanual.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/c298fb562803396052f172a4.pdf"},{"id":98730281,"identity":"d229150a-58b4-4fdc-b806-58b621b02f42","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":152537,"visible":true,"origin":"","legend":"","description":"","filename":"65c15fb96b234a90b3bacc932f28599d1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/3a6c2a57bdb0395886f8a897.xml"},{"id":98730285,"identity":"76fa1735-a550-4848-b6a6-89c8bbe04954","added_by":"auto","created_at":"2025-12-22 05:06:23","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":140698,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1/9efbea874aeded81178ff819.html"},{"id":98783628,"identity":"a615e448-370b-420e-95c8-6b636fc16466","added_by":"auto","created_at":"2025-12-22 12:42:26","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":988173,"visible":true,"origin":"","legend":"","description":"","filename":"ProtoMFL.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8376153/v1_covered_1dcda60d-1152-4da8-b982-6bdff4561647.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"ProtoMFL: A Robust Multimodal Federated Learning Framework via Cross-Modal Prototype Integration","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"the-journal-of-supercomputing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [The Journal of Supercomputing](https://www.springer.com/journal/11227)","snPcode":"11227","submissionUrl":"https://submission.nature.com/new-submission/11227/3","title":"The Journal of Supercomputing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Multimodal federated learning, modality missing, Prototype Learning, semantic alignment","lastPublishedDoi":"10.21203/rs.3.rs-8376153/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8376153/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMultimodal federated learning (MFL) has made substantial progress in aggregating multimodal knowledge across distributed environments. However, it still encounters persistent challenges caused by modality-missing data at the client level. Traditional knowledge distillation\u0026ndash;based approaches provide limited performance in handling these modality-missing scenarios. To mitigate the performance degradation caused by modality dropout, this paper proposes a prototype-based multimodal federated learning framework, termed Prototype-based Multimodal Federated Learning (ProtoMFL). By replacing sample-level representations with category-level prototypes as knowledge carriers, ProtoMFL enables more efficient cross-modal knowledge aggregation. The ProtoMFL framework consists of three core components. Cross-Modal Prototype Regularisation reduces distributional discrepancies between client and global models. Cross-Modal Prototype Contrast enhances the aggregation of similar prototypes and separation of dissimilar ones through contrastive learning. Cross-Modal Alignment enforces semantic alignment between modalities at the feature level, thereby mitigating the adverse effects of modality dropout. Experimental results show that ProtoMFL significantly outperforms existing methods in both accuracy and robustness across multiple benchmark datasets. Even under severe modality dropout, ProtoMFL maintains stable performance, achieving an average improvement of approximately 2.8% over the baseline CreamFL model without prototype mechanisms. This improvement effectively mitigates model drift issues caused by heterogeneous modalities.\u003c/p\u003e","manuscriptTitle":"ProtoMFL: A Robust Multimodal Federated Learning Framework via Cross-Modal Prototype Integration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-22 05:06:18","doi":"10.21203/rs.3.rs-8376153/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-11T17:29:32+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-08T14:20:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"143458154395887804474702325485569318914","date":"2025-12-27T10:02:26+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-25T17:04:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"307580482739185457021162779598322043809","date":"2025-12-21T13:40:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"126389340565468666557795035451297898173","date":"2025-12-19T14:39:32+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-18T04:00:03+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-17T02:39:57+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-17T02:39:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"The Journal of Supercomputing","date":"2025-12-16T12:30:20+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"the-journal-of-supercomputing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [The Journal of Supercomputing](https://www.springer.com/journal/11227)","snPcode":"11227","submissionUrl":"https://submission.nature.com/new-submission/11227/3","title":"The Journal of Supercomputing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"259e7648-1e67-4a42-93b5-57cf1c5ad223","owner":[],"postedDate":"December 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-04T22:38:43+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-22 05:06:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8376153","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8376153","identity":"rs-8376153","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}