ViT-CAAC: Contribution-Aware Adaptive Compression Framework for Vision Transformers

ViT-CAAC: Contribution-Aware Adaptive Compression Framework for Vision Transformers | 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 Article ViT-CAAC: Contribution-Aware Adaptive Compression Framework for Vision Transformers YU ZHANG, Shujun Peng, Yuheng Xiao, Xinhan Lin, Yang Hu, Shouyi Yin This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7464053/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 12 You are reading this latest preprint version Abstract The Vision Transformer (ViT) model has emerged as a powerful architecture for visual tasks by enabling the capture of long-range dependencies within images, demonstrating superior performance across a variety of applications. However, the large parameter count, along with high computational and memory demands of ViTs pose significant challenges. This paper introduces ViT-CAAC (Contribution-Aware Adaptive Compression Framework), a novel, multi-faceted compression framework designed to optimize ViTs. Our framework integrates block-level knowledge distillation, layer-wise quantization with precision control across hierarchical layers, and adaptive sparsity, creating a cohesive approach that substantially reduces model size while preserving performance. Through rigorous experimentation on benchmark datasets, we demonstrate that our framework achieves over 76% reduction in model size with minimal accuracy degradation (less than 0.4% Top-1 accuracy loss). This work establishes a novel concept for deploying high-performance vision models on resource-limited devices, with implications for applications in autonomous systems, IoT, and real-time vision processing. Physical sciences/Engineering Physical sciences/Mathematics and computing Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 16 Sep, 2025 Reviews received at journal 16 Sep, 2025 Reviews received at journal 15 Sep, 2025 Reviews received at journal 15 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers agreed at journal 07 Sep, 2025 Reviewers agreed at journal 07 Sep, 2025 Reviewers invited by journal 07 Sep, 2025 Editor invited by journal 29 Aug, 2025 Editor assigned by journal 29 Aug, 2025 Submission checks completed at journal 28 Aug, 2025 First submitted to journal 26 Aug, 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-7464053","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":512777812,"identity":"262eec65-f8da-4834-8484-d5a192dd409c","order_by":0,"name":"YU ZHANG","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"YU","middleName":"","lastName":"ZHANG","suffix":""},{"id":512777813,"identity":"301d2160-a609-4ef0-9ed9-8e9e46aeb147","order_by":1,"name":"Shujun Peng","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Shujun","middleName":"","lastName":"Peng","suffix":""},{"id":512777814,"identity":"9b2dce7e-21e4-4e5a-b433-25865aba78f1","order_by":2,"name":"Yuheng Xiao","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Yuheng","middleName":"","lastName":"Xiao","suffix":""},{"id":512777815,"identity":"6191bba3-9e25-4274-87b7-c77c445e9b2a","order_by":3,"name":"Xinhan Lin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIiWNgGAWjYBACPmYgkcBwgIGBvYHB4AGIQQiwwbXwHGAwSCBKC4QCqpRIgOolqIWd95jEwx135M0l3x4oSKi5w8AvffwCw88d+BzGl2yQeOaZ4c7ZeQkGCceeMUj25RQw9p7Bp4XH8EFi22HGDbdzDAwS2A4zGJzhSWBmbMOrxeAAUIv9hptngFr+EacFbEvihhs8BgZABlAL+wFCWoxBKpM3nAH6JbHvGY9kDw/DwV48Wvj5z5hJ/mw7bLvh+NljBh++3ZHj52F/+OAnHi1IgIfNAEQCkcEBojQAVTI/gDDYHxCpYxSMglEwCkYIAAAHuVNGEmKzVAAAAABJRU5ErkJggg==","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":true,"prefix":"","firstName":"Xinhan","middleName":"","lastName":"Lin","suffix":""},{"id":512777816,"identity":"fb636b8b-8abb-4c00-8e76-6c0362864b37","order_by":4,"name":"Yang Hu","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Hu","suffix":""},{"id":512777817,"identity":"5f2788d6-36a6-4a6b-b941-66439b8ff46b","order_by":5,"name":"Shouyi Yin","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Shouyi","middleName":"","lastName":"Yin","suffix":""}],"badges":[],"createdAt":"2025-08-26 14:38:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7464053/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7464053/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91212774,"identity":"f44a86d7-c554-4fb4-aaf3-7d8ad40fef25","added_by":"auto","created_at":"2025-09-12 18:33:15","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":539961,"visible":true,"origin":"","legend":"","description":"","filename":"ViTCAACtoScientificReports2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7464053/v1_covered_1440309d-ed04-48a4-85f7-5ed2551bd631.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"ViT-CAAC: Contribution-Aware Adaptive Compression Framework for Vision Transformers","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":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7464053/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7464053/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The Vision Transformer (ViT) model has emerged as a powerful architecture for visual tasks by enabling the capture of long-range dependencies within images, demonstrating superior performance across a variety of applications. However, the large parameter count, along with high computational and memory demands of ViTs pose significant challenges. This paper introduces ViT-CAAC (Contribution-Aware Adaptive Compression Framework), a novel, multi-faceted compression framework designed to optimize ViTs. Our framework integrates block-level knowledge distillation, layer-wise quantization with precision control across hierarchical layers, and adaptive sparsity, creating a cohesive approach that substantially reduces model size while preserving performance. Through rigorous experimentation on benchmark datasets, we demonstrate that our framework achieves over 76% reduction in model size with minimal accuracy degradation (less than 0.4% Top-1 accuracy loss). This work establishes a novel concept for deploying high-performance vision models on resource-limited devices, with implications for applications in autonomous systems, IoT, and real-time vision processing.","manuscriptTitle":"ViT-CAAC: Contribution-Aware Adaptive Compression Framework for Vision Transformers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-12 18:25:10","doi":"10.21203/rs.3.rs-7464053/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-16T19:13:35+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-16T15:59:54+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-15T05:31:36+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-15T04:47:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"97545957811201435695616097503109276584","date":"2025-09-10T01:27:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"168123787602872722339043695162901528052","date":"2025-09-08T00:30:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"324282953271666330320478638633033408390","date":"2025-09-08T00:28:07+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-08T00:20:34+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-08-29T20:28:16+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-29T06:46:47+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-28T14:31:20+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-08-26T14:30:50+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"86d231d6-1a55-4e19-81df-147f40ab0f59","owner":[],"postedDate":"September 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":54469334,"name":"Physical sciences/Engineering"},{"id":54469335,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2025-11-07T11:23:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-12 18:25:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7464053","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7464053","identity":"rs-7464053","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}