TransMamba:A language model combining Transformer and Mamba

TransMamba:A language model combining Transformer and Mamba | 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 TransMamba:A language model combining Transformer and Mamba Xiaocui Zhu, Qunsheng Ruan, Qian Sai, Miaohui Zhang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4782985/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Apr, 2025 Read the published version in Scientific Reports → Version 1 posted 21 You are reading this latest preprint version Abstract In recent years, State Space Models (SSMs) have achieved significant advancements in the field oflanguage modeling. With the advent of Mamba, these models have garnered even greater attention,surpassing Transformers in certain aspects. Despite Mamba’s unique advantages, Transformers remainindispensable due to their complex computational capabilities and proven effectiveness. This paperproposes a novel model that effectively combines the strengths of both Transformers and Mamba.Specifically, our model employs the Transformer’s encoder for encoding and utilizes Mamba as thedecoder for decoding. We introduce a feature fusion technique that integrates the features generated bythe encoder with the hidden states produced by the decoder. This approach effectively amalgamatesthe advantages of both Transformer and Mamba, resulting in enhanced performance. Extensiveexperiments on various language tasks demonstrate that our proposed model achieves competitiveresults, consistently outperforming existing benchmarks. Physical sciences/Mathematics and computing/Computer science Physical sciences/Mathematics and computing/Computational science State Space Models (SSMs) Transformer Mamba Feature Fusion Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 03 Apr, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 15 Oct, 2024 Reviews received at journal 15 Oct, 2024 Reviews received at journal 13 Oct, 2024 Reviewers agreed at journal 09 Oct, 2024 Reviewers agreed at journal 09 Oct, 2024 Reviewers agreed at journal 09 Oct, 2024 Reviewers agreed at journal 08 Oct, 2024 Reviews received at journal 08 Oct, 2024 Reviewers agreed at journal 08 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers agreed at journal 07 Oct, 2024 Reviewers invited by journal 07 Oct, 2024 Editor assigned by journal 07 Oct, 2024 Editor invited by journal 05 Aug, 2024 Submission checks completed at journal 29 Jul, 2024 First submitted to journal 22 Jul, 2024 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-4782985","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":343866603,"identity":"de1f2746-e406-47c1-aba3-faeb6ba9844f","order_by":0,"name":"Xiaocui Zhu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIie3RsQrCMBCA4SuBTFHXiGBe4YLg1IdJETo5CK6lCIIu4tyiD9HJuRLQpeDq2MlZEYqT2OrkEusmmB8CGe4jJAGw2X4xXi5VbQhJ8VtC1RfkFaspxGrR53ngCtyz85gFIbTmqXMdGYizznqodr5MdCOR0U4DzxTpRAZC+FDmimqFJcEzTQGOQAkzEMqHmKp7RViO6h6C+ERYdYo3exLAy4wAfiKc+2P0lr6MNUUZLzWTmTftmIiIBpv2rXBF86BPPVaE3e5eb68m8nav6mPKYWdSE5Svl9cetdlstr/qAcCgRGilg7ZzAAAAAElFTkSuQmCC","orcid":"","institution":"Jianxi Academy Sciences","correspondingAuthor":true,"prefix":"","firstName":"Xiaocui","middleName":"","lastName":"Zhu","suffix":""},{"id":343866604,"identity":"b79e567e-8449-4a8c-b914-c28348cadb4d","order_by":1,"name":"Qunsheng Ruan","email":"","orcid":"","institution":"Ganzhou Teachers College","correspondingAuthor":false,"prefix":"","firstName":"Qunsheng","middleName":"","lastName":"Ruan","suffix":""},{"id":343866605,"identity":"f16ec8ff-b705-401b-8fbf-412f07c9778c","order_by":2,"name":"Qian Sai","email":"","orcid":"","institution":"Jianxi Academy Sciences","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Sai","suffix":""},{"id":343866606,"identity":"3efaeb66-0488-4e76-b40e-4d4d6aed5a90","order_by":3,"name":"Miaohui Zhang","email":"","orcid":"","institution":"Jianxi Academy Sciences","correspondingAuthor":false,"prefix":"","firstName":"Miaohui","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2024-07-22 15:39:32","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4782985/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4782985/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-87574-8","type":"published","date":"2025-04-03T15:57:46+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":80082072,"identity":"fdb51d2e-5707-4306-8dbb-909e0d2cf091","added_by":"auto","created_at":"2025-04-07 16:06:40","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":560431,"visible":true,"origin":"","legend":"","description":"","filename":"TransMambaXiaocui.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4782985/v1_covered_9d031d0c-a9ff-42db-be3b-b2ae74620bc3.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"TransMamba:A language model combining Transformer and Mamba","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":"State Space Models (SSMs), Transformer, Mamba, Feature Fusion","lastPublishedDoi":"10.21203/rs.3.rs-4782985/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4782985/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"In recent years, State Space Models (SSMs) have achieved significant advancements in the field oflanguage modeling. With the advent of Mamba, these models have garnered even greater attention,surpassing Transformers in certain aspects. Despite Mamba’s unique advantages, Transformers remainindispensable due to their complex computational capabilities and proven effectiveness. This paperproposes a novel model that effectively combines the strengths of both Transformers and Mamba.Specifically, our model employs the Transformer’s encoder for encoding and utilizes Mamba as thedecoder for decoding. We introduce a feature fusion technique that integrates the features generated bythe encoder with the hidden states produced by the decoder. This approach effectively amalgamatesthe advantages of both Transformer and Mamba, resulting in enhanced performance. Extensiveexperiments on various language tasks demonstrate that our proposed model achieves competitiveresults, consistently outperforming existing benchmarks.","manuscriptTitle":"TransMamba:A language model combining Transformer and Mamba","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-23 05:24:35","doi":"10.21203/rs.3.rs-4782985/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-10-15T10:44:05+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-15T07:43:08+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-13T06:08:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"107460267869932911622145920082116366907","date":"2024-10-09T21:33:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"104847622640666393748781438858335845509","date":"2024-10-09T13:17:20+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"51002510983043028957953938807217413779","date":"2024-10-09T13:12:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"181973449515981749390208272748199929807","date":"2024-10-08T18:23:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-08T05:47:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"81175414924447361835983126398020981279","date":"2024-10-08T04:43:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"261852880558087652573988035120545048545","date":"2024-10-07T21:42:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"168509942703529541225660182119837352900","date":"2024-10-07T17:43:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"1306856300304455791055689892304209395","date":"2024-10-07T16:56:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"18457189418220345591763653501451642083","date":"2024-10-07T15:05:11+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"312259341908292864868024439285316658034","date":"2024-10-07T13:56:31+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"194077693883971475719463270517271074069","date":"2024-10-07T13:22:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"255461931742602893555469604744785792417","date":"2024-10-07T13:13:33+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-10-07T13:04:50+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-10-07T11:53:54+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-08-05T17:35:27+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-07-29T09:50:00+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-07-22T15:38:08+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":"c132ec86-8615-4b95-9be9-b1f54416f936","owner":[],"postedDate":"August 23rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":36442441,"name":"Physical sciences/Mathematics and computing/Computer science"},{"id":36442442,"name":"Physical sciences/Mathematics and computing/Computational science"}],"tags":[],"updatedAt":"2025-04-07T16:02:01+00:00","versionOfRecord":{"articleIdentity":"rs-4782985","link":"https://doi.org/10.1038/s41598-025-87574-8","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-04-03 15:57:46","publishedOnDateReadable":"April 3rd, 2025"},"versionCreatedAt":"2024-08-23 05:24:35","video":"","vorDoi":"10.1038/s41598-025-87574-8","vorDoiUrl":"https://doi.org/10.1038/s41598-025-87574-8","workflowStages":[]},"version":"v1","identity":"rs-4782985","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4782985","identity":"rs-4782985","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}