Enhancing Multimodal Recommendation via Contrastive Self-Supervised Modality-Preserving Learning

Enhancing Multimodal Recommendation via Contrastive Self-Supervised Modality-Preserving Learning | 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 Enhancing Multimodal Recommendation via Contrastive Self-Supervised Modality-Preserving Learning Jiajie Lu, Yamashita Haruka This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7555258/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Multimodal recommendation systems have gained increasing attention for their ability to incorporate rich side information such as visual and textual features. However, a critical yet underexplored challenge is the insufficient preservation of modality-specific information during training, which can weaken the effectiveness of multimodal signals and limit recommendation accuracy. To address this limitation, we propose Contrastive Modality-Preserving Learning (CMPL), a novel framework that extends the state-of-the-art MONET architecture. CMPL introduces a before-and-after contrastive learning module that explicitly maximizes the mutual information between initial modality embeddings and their final representations, thereby ensuring stronger modality preservation. At the same time, a graph convolutional backbone captures high-order collaborative signals from the user–item interaction graph, while a target-aware attention mechanism adaptively emphasizes user preference patterns. This joint design allows CMPL to balance the preservation of modality cues with the exploitation of collaborative filtering signals. We conduct extensive experiments on two real-world Amazon datasets, Office and MenClothing, and results consistently show that CMPL outperforms competitive baselines, including MARIO and MONET, in terms of precision and recall. These findings validate both the effectiveness of our approach and further highlight the necessity of explicitly modeling modality preservation for robust multimodal recommendation. Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 16 Mar, 2026 Reviews received at journal 16 Mar, 2026 Reviewers agreed at journal 08 Mar, 2026 Reviews received at journal 28 Dec, 2025 Reviewers agreed at journal 17 Dec, 2025 Reviewers invited by journal 12 Oct, 2025 Editor assigned by journal 07 Oct, 2025 Submission checks completed at journal 09 Sep, 2025 First submitted to journal 07 Sep, 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-7555258","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":533456635,"identity":"faf89160-3459-4672-9e71-99062dc9694e","order_by":0,"name":"Jiajie Lu","email":"","orcid":"","institution":"Sophia University","correspondingAuthor":false,"prefix":"","firstName":"Jiajie","middleName":"","lastName":"Lu","suffix":""},{"id":533456637,"identity":"05316238-e4f9-4933-8997-5ac0e9856923","order_by":1,"name":"Yamashita Haruka","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABMklEQVRIie3QsUrDQBgH8H8IXB2CrpGAeYULgRYx4qvkEC5LUgTBqYRAIV1iXLP1FQK+QCAQF3EuOHiTk4M4lKCleKkiHdLgKJj/8N1xd7/j7gP69PmDod+jSzYVzpEciJx8rXaTxghw+/dEnoMiULKoIV0ZQbPeFpNwvG9OBdx37s1n15UQE5gHEZ5FCzmONNvwq/KSEELBUifI7x886lawsgIebSH06YUaPilYTDCqWcKDXPeHuuyGkgNcbyOFZn/461CSwStYUnpm1pA1zjrI0AhiVRKNwq1LFwtJWAy2m5CrkyAtJfEvwCJuyb9wnaX6eVbu+Euh3j76y5DNp3e5Uq8c05QdO6yXzunNLOFtHQP2tm5S4qZqTZFPUjXeKjDYvmn1QzZbVTvp06dPn3+WT/GYZKKMRDviAAAAAElFTkSuQmCC","orcid":"","institution":"Sophia University","correspondingAuthor":true,"prefix":"","firstName":"Yamashita","middleName":"","lastName":"Haruka","suffix":""}],"badges":[],"createdAt":"2025-09-07 09:23:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7555258/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7555258/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94438791,"identity":"ac0dc016-d22a-4ecc-87a9-fe1f42933f53","added_by":"auto","created_at":"2025-10-27 14:21:57","extension":"png","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":384288,"visible":true,"origin":"","legend":"","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/5b718db71f21129dfd16a5c6.png"},{"id":94437891,"identity":"863e9b98-561c-42ef-add2-b00c3c6b7de3","added_by":"auto","created_at":"2025-10-27 14:21:11","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49697,"visible":true,"origin":"","legend":"","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/47c35f761a7e214641e94a16.png"},{"id":94438780,"identity":"10c4868f-7a63-4b46-8fed-051e74d73779","added_by":"auto","created_at":"2025-10-27 14:21:55","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":374588,"visible":true,"origin":"","legend":"","description":"","filename":"submissiondouble.docx","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/1bfd3ae5aa0a10111719e764.docx"},{"id":94438783,"identity":"1a568350-249d-4d61-9527-95d741fdc7aa","added_by":"auto","created_at":"2025-10-27 14:21:56","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11664,"visible":true,"origin":"","legend":"","description":"","filename":"tables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/cb5a0bf0fb4daec3bb69e464.xlsx"},{"id":94438459,"identity":"cebdda12-b45c-47dc-872f-a0aa2a03bee5","added_by":"auto","created_at":"2025-10-27 14:21:32","extension":"json","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4575,"visible":true,"origin":"","legend":"","description":"","filename":"714ee3c62e2f454e805beb646e998240.json","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/5597b81638f17553c700a574.json"},{"id":94489576,"identity":"98d25fb9-2072-4776-a300-e5e4627207e3","added_by":"auto","created_at":"2025-10-27 17:05:11","extension":"xml","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83890,"visible":true,"origin":"","legend":"","description":"","filename":"714ee3c62e2f454e805beb646e9982401enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/e9d1204bbe5c23561ed1f021.xml"},{"id":94438790,"identity":"4e36b3fe-e13e-4a6c-b429-3efa71a305f9","added_by":"auto","created_at":"2025-10-27 14:21:57","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":384288,"visible":true,"origin":"","legend":"","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/a37029e4bea0ea6c97b89851.png"},{"id":94438564,"identity":"477d9a5a-4bfc-4358-b2f6-c94b8042da10","added_by":"auto","created_at":"2025-10-27 14:21:32","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49697,"visible":true,"origin":"","legend":"","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/9b1209af75672302a2688ea4.png"},{"id":94438418,"identity":"cf39f481-7bb0-493f-80b3-e689200ead1f","added_by":"auto","created_at":"2025-10-27 14:21:31","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":266201,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/d1aa796a1fbb23f3e94555e9.png"},{"id":94438638,"identity":"c7dc172c-25d5-45f0-b3e2-f6b73b5720e4","added_by":"auto","created_at":"2025-10-27 14:21:38","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49697,"visible":true,"origin":"","legend":"","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/1a8dc61990ae605140794768.png"},{"id":94437657,"identity":"cfa51b37-24ff-4eb1-ac21-302b72c8386f","added_by":"auto","created_at":"2025-10-27 14:21:03","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":105212,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefig1.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/d2f4ab4fa4af6a45d38c90cb.png"},{"id":94438417,"identity":"78fa91ca-0d6b-4b9d-8090-cc62b591d152","added_by":"auto","created_at":"2025-10-27 14:21:31","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12849,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefig2.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/cded1d10fbaf4fc019d44d63.png"},{"id":94438806,"identity":"621244be-2f72-4f11-ab69-6081561e32bb","added_by":"auto","created_at":"2025-10-27 14:22:01","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70346,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/dd37cb904159653f86fec696.png"},{"id":94438755,"identity":"5c8162be-52ad-487d-81fa-944910d9d969","added_by":"auto","created_at":"2025-10-27 14:21:46","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12849,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/c0bb399b331153e67b5ea2a6.png"},{"id":94438759,"identity":"35477226-56c6-47de-9a23-781a9251eb18","added_by":"auto","created_at":"2025-10-27 14:21:47","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82413,"visible":true,"origin":"","legend":"","description":"","filename":"714ee3c62e2f454e805beb646e9982401structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/e9f4e058a2d40ca2b2600c4d.xml"},{"id":94438774,"identity":"52bdb394-7180-4ba6-8e2f-163b0bfc3c4e","added_by":"auto","created_at":"2025-10-27 14:21:52","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":95803,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1/94051ececf856d9d00e485d6.html"},{"id":94491286,"identity":"47b5cdfa-92b3-4940-bb45-1d0a82921324","added_by":"auto","created_at":"2025-10-27 17:24:16","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":557348,"visible":true,"origin":"","legend":"","description":"","filename":"submissiondouble.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7555258/v1_covered_3f244ccd-6134-4a99-bca7-c0f160cbd717.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Multimodal Recommendation via Contrastive Self-Supervised Modality-Preserving Learning","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"international-journal-of-multimedia-information-retrieval","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmir","sideBox":"Learn more about [International Journal of Multimedia Information Retrieval](http://link.springer.com/journal/13735)","snPcode":"13735","submissionUrl":"https://submission.nature.com/new-submission/13735/3","title":"International Journal of Multimedia Information Retrieval","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7555258/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7555258/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMultimodal recommendation systems have gained increasing attention for their ability to incorporate rich side information such as visual and textual features. However, a critical yet underexplored challenge is the insufficient preservation of modality-specific information during training, which can weaken the effectiveness of multimodal signals and limit recommendation accuracy. To address this limitation, we propose Contrastive Modality-Preserving Learning (CMPL), a novel framework that extends the state-of-the-art MONET architecture. CMPL introduces a before-and-after contrastive learning module that explicitly maximizes the mutual information between initial modality embeddings and their final representations, thereby ensuring stronger modality preservation. At the same time, a graph convolutional backbone captures high-order collaborative signals from the user\u0026ndash;item interaction graph, while a target-aware attention mechanism adaptively emphasizes user preference patterns. This joint design allows CMPL to balance the preservation of modality cues with the exploitation of collaborative filtering signals. We conduct extensive experiments on two real-world Amazon datasets, Office and MenClothing, and results consistently show that CMPL outperforms competitive baselines, including MARIO and MONET, in terms of precision and recall. These findings validate both the effectiveness of our approach and further highlight the necessity of explicitly modeling modality preservation for robust multimodal recommendation.\u003c/p\u003e","manuscriptTitle":"Enhancing Multimodal Recommendation via Contrastive Self-Supervised Modality-Preserving Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-27 10:41:19","doi":"10.21203/rs.3.rs-7555258/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-16T23:47:13+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-16T16:55:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"258978773854471684796581302214483427256","date":"2026-03-08T13:44:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-29T02:25:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"92230103378388242114568431772654998288","date":"2025-12-17T06:05:06+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-12T14:10:21+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-07T11:18:20+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-09T11:51:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Multimedia Information Retrieval","date":"2025-09-07T09:17:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-multimedia-information-retrieval","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmir","sideBox":"Learn more about [International Journal of Multimedia Information Retrieval](http://link.springer.com/journal/13735)","snPcode":"13735","submissionUrl":"https://submission.nature.com/new-submission/13735/3","title":"International Journal of Multimedia Information Retrieval","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"51c8496b-c7a3-472b-af81-d2f02c3bca30","owner":[],"postedDate":"October 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-27T02:38:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-27 10:41:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7555258","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7555258","identity":"rs-7555258","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}