Neural Networks Discover Physical Principles from Observation with an Intuitive World Model

Neural Networks Discover Physical Principles from Observation with an Intuitive World Model | 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 Neural Networks Discover Physical Principles from Observation with an Intuitive World Model Xiangyu Zhu, Qu Tang, Haoyuan Zhang, Zhen Lei, Zhaoxiang Zhang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3744228/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Humans readily grasp physical principles via visual observation, underpinning our ability to model the world and reason about events. In contrast, AI systems often fail to extract consistent physical rules from video, relying on black-box mechanisms that produce implausible outcomes. We introduce the Intuitive World Model (IWM), designed to learn physical principles directly from vision, mimicking human intuition. Drawing inspiration from developmental psychology, IWM utilizes a modular architecture to decompose interactions into force-based dynamics. Through observing objects collide and interact via charges, it identifies intrinsic properties like mass and charge and learns representations analogous to Newton's and Coulomb's laws. This paradigm uniquely allows neural networks to provide explicit explanations for physical phenomena and conduct counterfactual reasoning consistent with learned physical laws. Our work demonstrates that a developmentally-inspired approach empowers AI to acquire explicit, interpretable physical knowledge from observation, mirroring a core aspect of human intelligence. Physical sciences/Engineering/Mechanical engineering Physical sciences/Mathematics and computing/Computer science Full Text Additional Declarations There is NO Competing Interest. Supplementary Files IWMNCEsup.pdf Extended Analysis and Experimental Details for the Intuitive World Model Cite Share Download PDF Status: Under Review 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-3744228","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":451184541,"identity":"33084798-6bdd-4477-a8c2-736ce97f7f4d","order_by":0,"name":"Xiangyu Zhu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDklEQVRIiWNgGAWjYDACZhhDgoHxAUMBsggRWpgNGAyI0QIHEgxsEmAthIDBceZnD7+2HWaQn91jVl1gcCefgZ35AcOPGgZ5cxxaJJvZzI1lgVoY55wxuz3D4JllAzObAWPPMQbDnQ3YtfAzM5hJSwK1MEvkmN3mMThsAPSLAQNvA0OCwQHsWtiY2b+BtbABtRRDtLB/YPyLRws/M4+Z5EegFh6gFmaIFh4DZny2SDbzlEkznEvnkZBIK5bmMXhmwMbMU3BY5piE4QYcWgzOH98m+aPMWk5+RvLGzzwVdwz4+Y9vfPimxkYely0gwMzLxsADZR9gYAOTwGjCBxh//IGz8Zg8CkbBKBgFIxYAABpOSVj06g73AAAAAElFTkSuQmCC","orcid":"","institution":"Institute of Automation, Chinese Academy of Sciences","correspondingAuthor":true,"prefix":"","firstName":"Xiangyu","middleName":"","lastName":"Zhu","suffix":""},{"id":451184542,"identity":"d511beb2-1ef6-428f-900b-8566ab61464f","order_by":1,"name":"Qu Tang","email":"","orcid":"","institution":"Institute of Automation, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Qu","middleName":"","lastName":"Tang","suffix":""},{"id":451184543,"identity":"95d257c7-545f-434c-b1f3-0bc5eda81a0d","order_by":2,"name":"Haoyuan Zhang","email":"","orcid":"","institution":"Institute of Automation, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Haoyuan","middleName":"","lastName":"Zhang","suffix":""},{"id":451184544,"identity":"ee1e6882-45be-48cb-978f-db90fe55c302","order_by":3,"name":"Zhen Lei","email":"","orcid":"","institution":"Institute of Automation, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Zhen","middleName":"","lastName":"Lei","suffix":""},{"id":451184545,"identity":"adbc8ec6-b875-4dcd-a9ac-ef457712c1e6","order_by":4,"name":"Zhaoxiang Zhang","email":"","orcid":"","institution":"Institute of Automation, Chinese Academy of Science","correspondingAuthor":false,"prefix":"","firstName":"Zhaoxiang","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2023-12-12 14:06:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3744228/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3744228/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82211389,"identity":"aae18103-61de-476e-9c66-74cbf074ec17","added_by":"auto","created_at":"2025-05-07 19:29:02","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1370013,"visible":true,"origin":"","legend":"Article File","description":"","filename":"IWMNCE.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3744228/v1_covered_0de29b02-d0a9-4832-98ca-61c146964a3a.pdf"},{"id":82211008,"identity":"37517b28-19a7-4c1b-b016-d3fa1a482a18","added_by":"auto","created_at":"2025-05-07 19:21:00","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2050623,"visible":true,"origin":"","legend":"Extended Analysis and Experimental Details for the Intuitive World Model","description":"","filename":"IWMNCEsup.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3744228/v1/f92d50a768c628a8d7c54d32.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Neural Networks Discover Physical Principles from Observation with an Intuitive World Model","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":"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":"","lastPublishedDoi":"10.21203/rs.3.rs-3744228/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3744228/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Humans readily grasp physical principles via visual observation, underpinning our ability to model the world and reason about events. In contrast, AI systems often fail to extract consistent physical rules from video, relying on black-box mechanisms that produce implausible outcomes. We introduce the Intuitive World Model (IWM), designed to learn physical principles directly from vision, mimicking human intuition. Drawing inspiration from developmental psychology, IWM utilizes a modular architecture to decompose interactions into force-based dynamics. Through observing objects collide and interact via charges, it identifies intrinsic properties like mass and charge and learns representations analogous to Newton's and Coulomb's laws. This paradigm uniquely allows neural networks to provide explicit explanations for physical phenomena and conduct counterfactual reasoning consistent with learned physical laws. Our work demonstrates that a developmentally-inspired approach empowers AI to acquire explicit, interpretable physical knowledge from observation, mirroring a core aspect of human intelligence.","manuscriptTitle":"Neural Networks Discover Physical Principles from Observation with an Intuitive World Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-07 19:20:56","doi":"10.21203/rs.3.rs-3744228/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-engineering","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"commseng","sideBox":"Learn more about [Communications Engineering](http://link.springer.com/journal/44172)","snPcode":"44172","submissionUrl":"https://mts-commseng.nature.com/cgi-bin/main.plex","title":"Communications Engineering","twitterHandle":"@commseng","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"bd82635d-58f7-448d-bec0-7cb697e24d95","owner":[],"postedDate":"May 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":48003791,"name":"Physical sciences/Engineering/Mechanical engineering"},{"id":48003792,"name":"Physical sciences/Mathematics and computing/Computer science"}],"tags":[],"updatedAt":"2026-03-22T04:50:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-07 19:20:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3744228","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3744228","identity":"rs-3744228","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}