Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research

Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research | 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 Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research Boris Bolliet, Licong Xu, Andy Nilipour, Sebastien Pierre, Erwan Allys, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9114794/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract We present an agent-driven approach to the construction of parameter inference pipelines for scientific data analysis. Our method leverages a multi-agent system, Cmbagent 1 (the analysis system of the AI scientist Denario 2 ), in which specialized agents collaborate to generate research ideas, write and execute code, evaluate results, and iteratively refine the overall pipeline. As a case study, we apply this approach to the FAIR Universe Weak Lensing Uncertainty Challenge, a competition under time constraints focused on robust cosmological parameter inference with realistic observational uncertainties. While the fully autonomous exploration initially did not reach expertlevel performance, the integration of human intervention enabled our agent-driven workflow to achieve a first-place result in the challenge. This demonstrates that semi-autonomous agentic systems can compete with, and in some cases surpass, expert solutions. We describe our workflow in detail, including both the autonomous and semi-autonomous exploration by Cmbagent. Our final inference pipeline utilizes parameter-efficient convolutional neural networks, likelihood calibration over a known parameter grid, and multiple regularization techniques. Our results suggest that agent-driven research workflows can provide a scalable framework to rapidly explore and construct pipelines for inference problems. Physical sciences/Astronomy and planetary science/Astronomy and astrophysics/Cosmology Physical sciences/Astronomy and planetary science/Astronomy and astrophysics/Computational astrophysics Full Text Additional Declarations There is NO Competing Interest. Cite Share Download PDF Status: Posted 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-9114794","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":610864589,"identity":"36eeca19-9791-4b24-9f7d-408a0d90d863","order_by":0,"name":"Boris Bolliet","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuElEQVRIiWNgGAWjYBACAwnmg4//VMD5CcRoYUs24DlDmhYeNQHeNlK0mEv3sDFIzrOT121gfviBsS2NsBbLOWePPTDclmy47QCbsQRjWw4RDruRl26QuO0A47YDDGYMjG0VxGjJMZM4OOeA/bYD7N+I1yLZ2HAAaBEPyBYiHGY551iyMcOx5ORth3mKJRLOEeF9c+nmg48Zauxstx1v3/jhQ1kyYS0IwMxAVKyMglEwCkbBKCAGAADRwDmPearr6gAAAABJRU5ErkJggg==","orcid":"","institution":"University of Cambridge","correspondingAuthor":true,"prefix":"","firstName":"Boris","middleName":"","lastName":"Bolliet","suffix":""},{"id":610864590,"identity":"c1a57ba1-326c-4278-82ac-d8eca82c57f4","order_by":1,"name":"Licong Xu","email":"","orcid":"","institution":"University of Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Licong","middleName":"","lastName":"Xu","suffix":""},{"id":610864591,"identity":"84879b77-3010-4cb1-9a36-ef7f59bec899","order_by":2,"name":"Andy Nilipour","email":"","orcid":"","institution":"Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Andy","middleName":"","lastName":"Nilipour","suffix":""},{"id":610864592,"identity":"732f1f7f-9484-4a5c-b138-3770ad2d1461","order_by":3,"name":"Sebastien Pierre","email":"","orcid":"","institution":"ENS","correspondingAuthor":false,"prefix":"","firstName":"Sebastien","middleName":"","lastName":"Pierre","suffix":""},{"id":610864593,"identity":"5bd84d2c-9f34-4270-94b0-a434fb2eae25","order_by":4,"name":"Erwan Allys","email":"","orcid":"","institution":"ENS","correspondingAuthor":false,"prefix":"","firstName":"Erwan","middleName":"","lastName":"Allys","suffix":""},{"id":610864594,"identity":"e21461ab-0881-4725-bfe2-586ee4792d32","order_by":5,"name":"Celia Lecat","email":"","orcid":"","institution":"CentraleSupélec","correspondingAuthor":false,"prefix":"","firstName":"Celia","middleName":"","lastName":"Lecat","suffix":""},{"id":610864595,"identity":"93f85a59-7e84-458f-8418-e3eed8b4f3b0","order_by":6,"name":"Biwei Dai","email":"","orcid":"","institution":"IAS","correspondingAuthor":false,"prefix":"","firstName":"Biwei","middleName":"","lastName":"Dai","suffix":""},{"id":610864596,"identity":"1acdb0e0-a870-461e-a3ee-6c3261f31ec3","order_by":7,"name":"Po-Wen Chang","email":"","orcid":"","institution":"NERSC","correspondingAuthor":false,"prefix":"","firstName":"Po-Wen","middleName":"","lastName":"Chang","suffix":""},{"id":610864597,"identity":"ef513f5c-52e5-4997-b129-10a5c1a60a53","order_by":8,"name":"Wahid Bhimji","email":"","orcid":"","institution":"NERSC","correspondingAuthor":false,"prefix":"","firstName":"Wahid","middleName":"","lastName":"Bhimji","suffix":""},{"id":610864598,"identity":"66ca4689-2a63-4509-903a-dede011c8dbb","order_by":9,"name":"Thomas Borret","email":"","orcid":"","institution":"Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Borret","suffix":""}],"badges":[],"createdAt":"2026-03-13 12:36:43","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9114794/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9114794/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109079894,"identity":"fc0a58a6-6dfd-47ce-8162-56e237e9667a","added_by":"auto","created_at":"2026-05-12 11:22:27","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1053955,"visible":true,"origin":"","legend":"Article File","description":"","filename":"FAIRUniverseJournalTBS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9114794/v1_covered_dd50d1f3-7d92-4d12-9564-de47e72f54dc.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9114794/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9114794/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWe present an agent-driven approach to the construction of parameter inference pipelines for scientific data analysis. Our method leverages a multi-agent system, Cmbagent\u003csup\u003e1\u003c/sup\u003e (the analysis system of the AI scientist Denario\u003csup\u003e2\u003c/sup\u003e ), in which specialized agents collaborate to generate research ideas, write and execute code, evaluate results, and iteratively refine the overall pipeline. As a case study, we apply this approach to the FAIR Universe Weak Lensing Uncertainty Challenge, a competition under time constraints focused on robust cosmological parameter inference with realistic observational uncertainties. While the fully autonomous exploration initially did not reach expertlevel performance, the integration of human intervention enabled our agent-driven workflow to achieve a first-place result in the challenge. This demonstrates that semi-autonomous agentic systems can compete with, and in some cases surpass, expert solutions. We describe our workflow in detail, including both the autonomous and semi-autonomous exploration by Cmbagent. Our final inference pipeline utilizes parameter-efficient convolutional neural networks, likelihood calibration over a known parameter grid, and multiple regularization techniques. Our results suggest that agent-driven research workflows can provide a scalable framework to rapidly explore and construct pipelines for inference problems.\u003c/p\u003e","manuscriptTitle":"Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-25 06:27:14","doi":"10.21203/rs.3.rs-9114794/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"bd947b97-0af6-446d-adab-eafaf51f886e","owner":[],"postedDate":"March 25th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Reject after peer review","date":"2026-05-12T10:21:47+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-05-04T12:54:56+00:00","index":2,"fulltext":"This content is not available."}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64997021,"name":"Physical sciences/Astronomy and planetary science/Astronomy and astrophysics/Cosmology"},{"id":64997022,"name":"Physical sciences/Astronomy and planetary science/Astronomy and astrophysics/Computational astrophysics"}],"tags":[],"updatedAt":"2026-05-12T11:11:18+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-25 06:27:14","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9114794","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9114794","identity":"rs-9114794","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}