Large Language Models as Materials Science Adapted Learners

Large Language Models as Materials Science Adapted Learners | 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 Large Language Models as Materials Science Adapted Learners Tong Xie, Yuwei Wan, Yixuan Liu, Yuchen Zeng, Shaozhou Wang, Wenjie Zhang, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6752901/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 9 You are reading this latest preprint version Abstract Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on complex descriptors, which hinder generalizability and transferability across different material systems. Moreover, these descriptors may inadequately represent macro-scale material properties, which are influenced by structural imperfections and compositional variations in real-world samples, thus limiting their practical applicability. To address these challenges, we propose DARWIN 1.5, the largest 1 open-source large language model tailored for materials science. By utilizing natural language as input, DARWIN eliminates the need for task-specific descriptors and facilitates the integration of human knowledge representation with computational models, enabling a more flexible and unified approach to material property prediction and discovery. Our approach integrates over 6M materials science papers and 21 experimental datasets with information of 49,256 materials, allowing for efficient cross-task knowledge transfer and improved generalization. Through systematic exploration, we show how domain-specific know-how can be effectively integrated into language models while harnessing the inherent syn-ergies between tasks to enhance predictive performance across diverse material science applications. The enhanced model achieves up to 59.1% improvement in prediction accuracy over the base LLaMA-7B model architecture and outper-forms state-of-the-art machine learning approaches across eight materials design tasks. These results highlight the potential of LLMs as a foundation for developing versatile and scalable models in materials science. Physical sciences/Chemistry/Theoretical chemistry Physical sciences/Materials science Physical sciences/Materials science/Techniques and instrumentation/Design synthesis and processing Large language models AI for Science material design material science Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 13 Sep, 2025 Reviews received at journal 09 Sep, 2025 Reviewers agreed at journal 29 Aug, 2025 Reviews received at journal 23 Jul, 2025 Reviewers agreed at journal 03 Jul, 2025 Reviewers invited by journal 03 Jul, 2025 Editor assigned by journal 02 Jun, 2025 Submission checks completed at journal 29 May, 2025 First submitted to journal 26 May, 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-6752901","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":480447997,"identity":"8a83fde1-d016-443a-983e-cd437b36807b","order_by":0,"name":"Tong Xie","email":"","orcid":"","institution":"GreenDynamics","correspondingAuthor":false,"prefix":"","firstName":"Tong","middleName":"","lastName":"Xie","suffix":""},{"id":480447998,"identity":"146243c2-8e59-4813-a2d4-9f78177c27b4","order_by":1,"name":"Yuwei Wan","email":"","orcid":"","institution":"GreenDynamics","correspondingAuthor":false,"prefix":"","firstName":"Yuwei","middleName":"","lastName":"Wan","suffix":""},{"id":480447999,"identity":"35d017c8-b37c-42a1-8647-ae53a530d046","order_by":2,"name":"Yixuan Liu","email":"","orcid":"","institution":"GreenDynamics","correspondingAuthor":false,"prefix":"","firstName":"Yixuan","middleName":"","lastName":"Liu","suffix":""},{"id":480448000,"identity":"d2a4ae92-cf3d-4f39-8365-ab5808a7b43d","order_by":3,"name":"Yuchen Zeng","email":"","orcid":"","institution":"University of Wisconsin-Madison","correspondingAuthor":false,"prefix":"","firstName":"Yuchen","middleName":"","lastName":"Zeng","suffix":""},{"id":480448001,"identity":"f4de25f3-e82d-431f-b155-7e443204c0d9","order_by":4,"name":"Shaozhou Wang","email":"","orcid":"","institution":"GreenDynamics","correspondingAuthor":false,"prefix":"","firstName":"Shaozhou","middleName":"","lastName":"Wang","suffix":""},{"id":480448002,"identity":"6b2b52a4-3ce6-49fa-8a71-055a07947f16","order_by":5,"name":"Wenjie Zhang","email":"","orcid":"","institution":"University of New South Wales","correspondingAuthor":false,"prefix":"","firstName":"Wenjie","middleName":"","lastName":"Zhang","suffix":""},{"id":480448003,"identity":"f3fc357b-1007-4b25-b791-2fbfa9f8c4e5","order_by":6,"name":"Clara Grazian","email":"","orcid":"","institution":"University of Sydney","correspondingAuthor":false,"prefix":"","firstName":"Clara","middleName":"","lastName":"Grazian","suffix":""},{"id":480448004,"identity":"29cfe537-f293-407e-a730-ea60f7e0854b","order_by":7,"name":"Chunyu Kit","email":"","orcid":"","institution":"City University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Chunyu","middleName":"","lastName":"Kit","suffix":""},{"id":480448005,"identity":"db6bd601-35ef-4910-8e9e-fe0e4c5153e4","order_by":8,"name":"Wanli Ouyang","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Wanli","middleName":"","lastName":"Ouyang","suffix":""},{"id":480448006,"identity":"3431ee1a-d613-4e5b-869c-1a72d9b7d32c","order_by":9,"name":"Dongzhan Zhou","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Dongzhan","middleName":"","lastName":"Zhou","suffix":""},{"id":480448007,"identity":"ff9fc6a1-1c27-44c6-a86d-400595dabe4d","order_by":10,"name":"Bram Hoex","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYBACAwYeEGUD5bIRryWNdC2HSdBizsB78MHHtvOJ2/nPGDB8KDvMwD8jAb8Wywa+ZMMZZ24n7pyRY8A449xhBokbBLQYHOAxk+apuJ244QbvBmbeNqALidLyx+Bc4obzZzcw/wVqkSdKC0PFgcQNB3I3MDMCtRgQ0mLZzGNs2HMm2XjDjfwPB3vOpfMYnnmAX4s5e4/hg59tdrIbzh9LfPCjzFpO7jgBWxiYkdgHgJiHgPpRMApGwSgYBcQAAI+KRN2wlyaoAAAAAElFTkSuQmCC","orcid":"","institution":"University of New South Wales","correspondingAuthor":true,"prefix":"","firstName":"Bram","middleName":"","lastName":"Hoex","suffix":""}],"badges":[],"createdAt":"2025-05-26 17:23:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6752901/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6752901/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":86156260,"identity":"6a663cc1-9939-49da-9cc1-8fe5eeec5db4","added_by":"auto","created_at":"2025-07-07 11:12:59","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4864660,"visible":true,"origin":"","legend":"","description":"","filename":"6114530artfile10767908svzzwc.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6752901/v1_covered_7bc9cb7e-85b6-4778-898a-e5a76f10391d.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Large Language Models as Materials Science Adapted Learners","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":"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":"Large language models, AI for Science, material design, material science","lastPublishedDoi":"10.21203/rs.3.rs-6752901/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6752901/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on complex descriptors, which hinder generalizability and transferability across different material systems. Moreover, these descriptors may inadequately represent macro-scale material properties, which are influenced by structural imperfections and compositional variations in real-world samples, thus limiting their practical applicability. To address these challenges, we propose DARWIN 1.5, the largest 1 open-source large language model tailored for materials science. By utilizing natural language as input, DARWIN eliminates the need for task-specific descriptors and facilitates the integration of human knowledge representation with computational models, enabling a more flexible and unified approach to material property prediction and discovery. Our approach integrates over 6M materials science papers and 21 experimental datasets with information of 49,256 materials, allowing for efficient cross-task knowledge transfer and improved generalization. Through systematic exploration, we show how domain-specific know-how can be effectively integrated into language models while harnessing the inherent syn-ergies between tasks to enhance predictive performance across diverse material science applications. The enhanced model achieves up to 59.1% improvement in prediction accuracy over the base LLaMA-7B model architecture and outper-forms state-of-the-art machine learning approaches across eight materials design tasks. These results highlight the potential of LLMs as a foundation for developing versatile and scalable models in materials science.","manuscriptTitle":"Large Language Models as Materials Science Adapted Learners","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-07 11:04:47","doi":"10.21203/rs.3.rs-6752901/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-13T14:15:32+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-09T04:44:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"117901143059750272564838632138592679455","date":"2025-08-29T04:23:04+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-24T01:56:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"259713382065813677354622502364993692190","date":"2025-07-03T23:30:56+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-03T12:30:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-02T09:10:24+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-29T07:46:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Computational Materials","date":"2025-05-26T17:11:33+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":"f97b6057-808b-41b8-a97c-cca53cfe451b","owner":[],"postedDate":"July 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":51021180,"name":"Physical sciences/Chemistry/Theoretical chemistry"},{"id":51021181,"name":"Physical sciences/Materials science"},{"id":51021182,"name":"Physical sciences/Materials science/Techniques and instrumentation/Design synthesis and processing"}],"tags":[],"updatedAt":"2025-09-13T14:23:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-07 11:04:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6752901","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6752901","identity":"rs-6752901","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}