LAMBench: A Benchmark for Large Atomistic Models

LAMBench: A Benchmark for Large Atomistic Models | 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 LAMBench: A Benchmark for Large Atomistic Models Anyang Peng, Chun Cai, Mingyu Guo, Duo Zhang, Chengqian Zhang, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7395628/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Jan, 2026 Read the published version in npj Computational Materials → Version 1 posted 8 You are reading this latest preprint version Abstract Large Atomistic Models (LAMs) have undergone remarkable progress recently, emerging as universal or fundamental representations of the potential energy surface defined by the first-principlescalculations of atomistic systems. However, our understanding of the extent to which these models achieve true universality, as well as their comparative performance across different models, remains limited. This gap is largely due to the lack of comprehensive benchmarks capable of evaluating the effectiveness of LAMs as approximations to the universal potential energy surface. In this study, we introduce LAMBench, a benchmarking system designed to evaluate LAMs in terms of their generalizability, adaptability, and applicability. These attributes are crucial for deploying LAMs as ready-to-use tools across a diverse array of scientific discovery contexts. We benchmark ten state-of-the-art LAMs released prior to August 1, 2025, using LAMBench. Our findings reveal asignificant gap between the current LAMs and the ideal universal potential energy surface. They also highlight the need for incorporating cross-domain training data, supporting multi-fidelity modeling, and ensuring the models’ conservativeness and differentiability. As a dynamic and extensible platform, LAMBench is intended to continuously evolve, thereby facilitating the development of robust and generalizable LAMs capable of significantly advancing scientific research. The LAMBench code is open-sourced at https://github.com/deepmodeling/lambench , and an interactive leaderboard is available at https://www.aissquare.com/openlam?tab=Benchmark . Physical sciences/Materials science Physical sciences/Mathematics and computing Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 03 Jan, 2026 Read the published version in npj Computational Materials → Version 1 posted Editorial decision: Revision requested 29 Oct, 2025 Reviews received at journal 28 Oct, 2025 Reviewers agreed at journal 17 Oct, 2025 Reviewers agreed at journal 02 Sep, 2025 Reviewers invited by journal 28 Aug, 2025 Editor assigned by journal 27 Aug, 2025 Submission checks completed at journal 26 Aug, 2025 First submitted to journal 18 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-7395628","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":509624981,"identity":"2584328b-0a29-47b9-9c81-ccc1d443b9dc","order_by":0,"name":"Anyang Peng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYPACGwY2BsYGkrSkka7lMAlqzdl7D7/42XZeno//cAPDjxoGeXNCWix7zqVZ9rbdNmyTSGxg7DnGYLiTkPsMbuSYGfC23WZskwD6hbeBIcHgACEt99+YGf5tO2ffxn+wgfEvUVpu8Bg/5m07kNjGkNjATJQtlj05Zswy55KTQX45LHNMwnADIS3m7GeMP74ps7Od33/84cM3NTbyhB3GwMAmwcgG4QAVSxBQD9HC/IHhD2GFo2AUjIJRMIIBADZ/P8YHUBlrAAAAAElFTkSuQmCC","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":true,"prefix":"","firstName":"Anyang","middleName":"","lastName":"Peng","suffix":""},{"id":509624982,"identity":"c3f31813-f3c6-4102-b337-d736465737de","order_by":1,"name":"Chun Cai","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Chun","middleName":"","lastName":"Cai","suffix":""},{"id":509624984,"identity":"a82cf3ba-59b3-4021-8c4c-3aeda446c92a","order_by":2,"name":"Mingyu Guo","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Mingyu","middleName":"","lastName":"Guo","suffix":""},{"id":509624985,"identity":"67fc6dcc-e646-4f85-ba42-9ef62f785661","order_by":3,"name":"Duo Zhang","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Duo","middleName":"","lastName":"Zhang","suffix":""},{"id":509624986,"identity":"13e923ce-ee42-4067-baf7-aef09301a26e","order_by":4,"name":"Chengqian Zhang","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Chengqian","middleName":"","lastName":"Zhang","suffix":""},{"id":509624987,"identity":"784d9267-bc2f-4e1b-8665-6ae32feebd93","order_by":5,"name":"Wanrun Jiang","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Wanrun","middleName":"","lastName":"Jiang","suffix":""},{"id":509624988,"identity":"409ae4c2-35e7-4041-ad13-911be571b32a","order_by":6,"name":"Yinan Wang","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Yinan","middleName":"","lastName":"Wang","suffix":""},{"id":509624990,"identity":"e9d5260e-e886-4e69-84fa-26200aa14a1f","order_by":7,"name":"Antoine Loew","email":"","orcid":"","institution":"Ruhr University Bochum","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"","lastName":"Loew","suffix":""},{"id":509624991,"identity":"9c32b1f3-a538-4257-a52e-6c170c02d784","order_by":8,"name":"Chengkun Wu","email":"","orcid":"","institution":"National University of Defense Technology","correspondingAuthor":false,"prefix":"","firstName":"Chengkun","middleName":"","lastName":"Wu","suffix":""},{"id":509624992,"identity":"8c240e8c-1511-4c96-8e5f-e1781422be0e","order_by":9,"name":"Weinan E","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Weinan","middleName":"","lastName":"E","suffix":""},{"id":509624993,"identity":"179c9c9e-f46a-405a-a42a-8cdb8f75a47c","order_by":10,"name":"Linfeng Zhang","email":"","orcid":"","institution":"AI for Science Institute, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Linfeng","middleName":"","lastName":"Zhang","suffix":""},{"id":509624996,"identity":"1ab2d0c3-caaf-43db-a64b-2ab542acc02d","order_by":11,"name":"Han Wang","email":"","orcid":"","institution":"Institute of Applied Physics and Computational Mathematics","correspondingAuthor":false,"prefix":"","firstName":"Han","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2025-08-18 04:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7395628/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7395628/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41524-025-01929-3","type":"published","date":"2026-01-03T15:58:37+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":99546138,"identity":"d064733b-697d-4a74-a1e4-9718e2d49b71","added_by":"auto","created_at":"2026-01-05 16:10:08","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2185707,"visible":true,"origin":"","legend":"","description":"","filename":"LamBench.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7395628/v1_covered_8e42f030-6a43-4410-b662-a0397d610b08.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"LAMBench: A Benchmark for Large Atomistic Models","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":"npj-computational-materials","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjcompumats","sideBox":"Learn more about [npj Computational Materials](http://www.nature.com/npjcompumats/)","snPcode":"41524","submissionUrl":"https://mts-npjcompumats.nature.com/","title":"npj Computational Materials","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7395628/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7395628/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Large Atomistic Models (LAMs) have undergone remarkable progress recently, emerging as universal or fundamental representations of the potential energy surface defined by the first-principlescalculations of atomistic systems. However, our understanding of the extent to which these models achieve true universality, as well as their comparative performance across different models, remains limited. This gap is largely due to the lack of comprehensive benchmarks capable of evaluating the effectiveness of LAMs as approximations to the universal potential energy surface. In this study, we introduce LAMBench, a benchmarking system designed to evaluate LAMs in terms of their generalizability, adaptability, and applicability. These attributes are crucial for deploying LAMs as ready-to-use tools across a diverse array of scientific discovery contexts. We benchmark ten state-of-the-art LAMs released prior to August 1, 2025, using LAMBench. Our findings reveal asignificant gap between the current LAMs and the ideal universal potential energy surface. They also highlight the need for incorporating cross-domain training data, supporting multi-fidelity modeling, and ensuring the models’ conservativeness and differentiability. As a dynamic and extensible platform, LAMBench is intended to continuously evolve, thereby facilitating the development of robust and generalizable LAMs capable of significantly advancing scientific research. The LAMBench code is open-sourced at https://github.com/deepmodeling/lambench, and an interactive leaderboard is available at https://www.aissquare.com/openlam?tab=Benchmark.","manuscriptTitle":"LAMBench: A Benchmark for Large Atomistic Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-05 03:03:46","doi":"10.21203/rs.3.rs-7395628/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-29T09:55:04+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-28T15:40:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"58191159856682733178048162194975453843","date":"2025-10-17T07:42:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"111211518687688959775802261218793200851","date":"2025-09-02T21:04:57+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-28T15:36:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-27T13:07:44+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-26T06:45:17+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Computational Materials","date":"2025-08-18T04:13:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-computational-materials","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjcompumats","sideBox":"Learn more about [npj Computational Materials](http://www.nature.com/npjcompumats/)","snPcode":"41524","submissionUrl":"https://mts-npjcompumats.nature.com/","title":"npj Computational Materials","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5ecf3042-af77-45d8-b6e3-0a024269aaac","owner":[],"postedDate":"September 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":54127880,"name":"Physical sciences/Materials science"},{"id":54127881,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-01-05T16:08:33+00:00","versionOfRecord":{"articleIdentity":"rs-7395628","link":"https://doi.org/10.1038/s41524-025-01929-3","journal":{"identity":"npj-computational-materials","isVorOnly":false,"title":"npj Computational Materials"},"publishedOn":"2026-01-03 15:58:37","publishedOnDateReadable":"January 3rd, 2026"},"versionCreatedAt":"2025-09-05 03:03:46","video":"","vorDoi":"10.1038/s41524-025-01929-3","vorDoiUrl":"https://doi.org/10.1038/s41524-025-01929-3","workflowStages":[]},"version":"v1","identity":"rs-7395628","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7395628","identity":"rs-7395628","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}