Online Test-time Adaptation for Interatomic Potentials

Online Test-time Adaptation for Interatomic Potentials | 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 Online Test-time Adaptation for Interatomic Potentials Shufei Zhang, Taoyong Cui, Chenyu Tang, Dongzhan Zhou, Yuqiang Li, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4388576/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 22 Feb, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Machine learning interatomic potentials (MLIPs) enable more efficient molecular dynamics (MD) simulations with ab initio accuracy, which have been used in various domains of physical science. However, distribution shift between training and test data causes deterioration of the test performance of MLIPs, and even leads to collapse of MD simulations. In this work, we propose an online Test-time Adaptation Interatomic Potential (TAIP) framework to improve the generalization on test data. Specifically, we design a dual-level self-supervised learning approach that leverages global structure and atomic local environment information to align the model with the test data. Extensive experiments demonstrate TAIP's capability to bridge the domain gap between training and test dataset without additional data. TAIP enhances the test performance on various benchmarks, from small molecule datasets to complex periodic molecular systems with various types of elements. Remarkably, it also enables stable MD simulations where the corresponding baseline models collapse. Physical sciences/Mathematics and computing/Computational science Physical sciences/Materials science/Theory and computation/Computational methods Test-time Adaptation Interatomic Potential Molecular Dynamics Simulation Full Text Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryinformationOnlineTesttimeAdaptationforInteratomicPotentials.pdf TheMDtrajectoriesoficeat300KwithPaiNNTAIP.mp4 The MD trajectories of ice at 300K with PaiNN-TAIP TheMDtrajectoriesofelectrolytesolution1MNaXTf2NXinECDMCwithat300KwithPaiNNTAIP.mp4 The MD trajectories of electrolyte solution (1M [NaX+][Tf_2NX-] in EC+DMC) with at 300K with PaiNN-TAIP Cite Share Download PDF Status: Published Journal Publication published 22 Feb, 2025 Read the published version in Nature Communications → 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-4388576","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":311781089,"identity":"f8b4f334-ecd8-4ed4-acb7-3ee88b26e817","order_by":0,"name":"Shufei Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIie3RsUrDQBjA8e8jkCzfA3whLXmFSKB10PoqHhm6FFEKRXCJHNx0D1CoD+GkdDJyUB+iDmbpXLFIhgxeCl2EGN0c7j8cdx/8hrsDcLn+Y7xfCwhyvH0DOobk14QKlAkQ/4mAz82xk8QLuSl39avQgVQ3lz2Oh+Ct3ncwumgjeLcapn21EZqe1XpOfLTM/WzRg2zaRjw+H0RhbsQTC7UmYrwvKPUYCpG3EJ/HnxHXRui4VFNLzjoJ8WQQbn1LGJVniWgIbn8gzJNZhMqkmoSM7F2ypfQzD5KslcTz8UNY1aavg5fy40qfnD4G0mB1PWol+yegwwZ1szaTjt/B6kCg+jZxuVwul+0LDL5VPNnrcf4AAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-4614-4718","institution":"Shanghai AI Laboratory","correspondingAuthor":true,"prefix":"","firstName":"Shufei","middleName":"","lastName":"Zhang","suffix":""},{"id":311781090,"identity":"a6e35f7d-67fb-4533-a6b7-5e0af80ff0ea","order_by":1,"name":"Taoyong Cui","email":"","orcid":"https://orcid.org/0000-0002-5532-6636","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Taoyong","middleName":"","lastName":"Cui","suffix":""},{"id":311781091,"identity":"377d0ad5-7a4b-43b6-a8ce-784b31e65e41","order_by":2,"name":"Chenyu Tang","email":"","orcid":"https://orcid.org/0000-0002-6914-7348","institution":"1. CAS Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy of Sciences. 2. School of Physical Sciences, University of Chinese Academy of Sciences.","correspondingAuthor":false,"prefix":"","firstName":"Chenyu","middleName":"","lastName":"Tang","suffix":""},{"id":311781092,"identity":"5551e398-b4ef-4e54-b369-5dd451e7dce0","order_by":3,"name":"Dongzhan Zhou","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Dongzhan","middleName":"","lastName":"Zhou","suffix":""},{"id":311781093,"identity":"0ffbcfe6-d588-49d3-bffd-8e24112fe8dd","order_by":4,"name":"Yuqiang Li","email":"","orcid":"https://orcid.org/0000-0001-6756-6154","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Yuqiang","middleName":"","lastName":"Li","suffix":""},{"id":311781094,"identity":"b00b0829-e5ab-443c-a390-1c5e94f2faed","order_by":5,"name":"Xingao Gong","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xingao","middleName":"","lastName":"Gong","suffix":""},{"id":311781095,"identity":"1c6c935f-27e9-460e-a790-dfcc166551e4","order_by":6,"name":"Wanli Ouyang","email":"","orcid":"","institution":"Shanghai AI lab","correspondingAuthor":false,"prefix":"","firstName":"Wanli","middleName":"","lastName":"Ouyang","suffix":""},{"id":311781096,"identity":"e04803c4-b822-4234-8c00-b962cf4792e4","order_by":7,"name":"Mao Su","email":"","orcid":"https://orcid.org/0000-0003-2066-881X","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Mao","middleName":"","lastName":"Su","suffix":""}],"badges":[],"createdAt":"2024-05-08 10:35:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4388576/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4388576/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-025-57101-4","type":"published","date":"2025-02-22T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":76951721,"identity":"9e363f0e-da7e-430e-a13b-6ea30794240e","added_by":"auto","created_at":"2025-02-23 08:06:51","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1654020,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineTesttimeAdaptationforInteratomicPotentials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4388576/v1_covered_e95c709d-9113-49f5-b961-fc6127cb3f6b.pdf"},{"id":58335408,"identity":"5bd769ea-3c17-466a-b7c1-df9263b542f8","added_by":"auto","created_at":"2024-06-14 05:19:51","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":7915719,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryinformationOnlineTesttimeAdaptationforInteratomicPotentials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4388576/v1/237be9a704543e8de043df31.pdf"},{"id":58335399,"identity":"ca880aab-5fc0-4090-9a06-dde9e63e6d6c","added_by":"auto","created_at":"2024-06-14 05:19:49","extension":"mp4","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":4683115,"visible":true,"origin":"","legend":"\u003cp\u003eThe MD trajectories of ice at 300K with PaiNN-TAIP\u003c/p\u003e","description":"","filename":"TheMDtrajectoriesoficeat300KwithPaiNNTAIP.mp4","url":"https://assets-eu.researchsquare.com/files/rs-4388576/v1/bf5d6c34d64e4309ed080cb5.mp4"},{"id":58335407,"identity":"88abc869-01f6-43fd-9675-1bd4f72af84b","added_by":"auto","created_at":"2024-06-14 05:19:51","extension":"mp4","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":6860853,"visible":true,"origin":"","legend":"\u003cp\u003eThe MD trajectories of electrolyte solution (1M [NaX+][Tf_2NX-] in EC+DMC) with at 300K with PaiNN-TAIP\u003c/p\u003e","description":"","filename":"TheMDtrajectoriesofelectrolytesolution1MNaXTf2NXinECDMCwithat300KwithPaiNNTAIP.mp4","url":"https://assets-eu.researchsquare.com/files/rs-4388576/v1/f199b8e74f2fa9c938e16666.mp4"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Online Test-time Adaptation for Interatomic\nPotentials","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Test-time Adaptation, Interatomic Potential, Molecular Dynamics Simulation","lastPublishedDoi":"10.21203/rs.3.rs-4388576/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4388576/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Machine learning interatomic potentials (MLIPs) enable more efficient molecular dynamics (MD) simulations with ab initio accuracy, which have been used in various domains of physical science. However, distribution shift between training and test data causes deterioration of the test performance of MLIPs, and even leads to collapse of MD simulations. In this work, we propose an online Test-time Adaptation Interatomic Potential (TAIP) framework to improve the generalization on test data. Specifically, we design a dual-level self-supervised learning approach that leverages global structure and atomic local environment information to align the model with the test data. Extensive experiments demonstrate TAIP's capability to bridge the domain gap between training and test dataset without additional data. TAIP enhances the test performance on various benchmarks, from small molecule datasets to complex periodic molecular systems with various types of elements. Remarkably, it also enables stable MD simulations where the corresponding baseline models collapse.","manuscriptTitle":"Online Test-time Adaptation for Interatomic\nPotentials","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-14 05:19:44","doi":"10.21203/rs.3.rs-4388576/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"6512d3e4-5dfc-43d8-b65e-8487e38ce245","owner":[],"postedDate":"June 14th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":32956338,"name":"Physical sciences/Mathematics and computing/Computational science"},{"id":32956339,"name":"Physical sciences/Materials science/Theory and computation/Computational methods"}],"tags":[],"updatedAt":"2025-02-23T08:06:43+00:00","versionOfRecord":{"articleIdentity":"rs-4388576","link":"https://doi.org/10.1038/s41467-025-57101-4","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2025-02-22 05:00:00","publishedOnDateReadable":"February 22nd, 2025"},"versionCreatedAt":"2024-06-14 05:19:44","video":"","vorDoi":"10.1038/s41467-025-57101-4","vorDoiUrl":"https://doi.org/10.1038/s41467-025-57101-4","workflowStages":[]},"version":"v1","identity":"rs-4388576","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4388576","identity":"rs-4388576","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}