A Multi-agent Court to Mitigate VLM Hallucinations

A Multi-agent Court to Mitigate VLM Hallucinations | 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 Research Article A Multi-agent Court to Mitigate VLM Hallucinations Percy Lam, Lavindra de Silva, Weiwei Chen, Ioannis Brilakis This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8949950/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 Hallucinations have hindered the widespread use of vision language models (VLMs) for domain-specific applications such as road maintenance. While previous researchers constructed multiple solutions for different sources of visual hallucinations, knowledge gaps persist in handling context-dependent hallucinations where the targeted objects are difficult to be prompted precisely. This research explores hallucination by converting image-to-text binary classifications into evidential arguments by VLM agents, each providing a binary Yes/No answer with a justification. The proposed solution involves VLM agents performing distinct roles, starting with a detection unit that uses a primary detector and a reviewer to verify scope compatibility. These agents interact to aggregate their findings and justifications into a single, unified verdict. The different roles of each agent are inspired by the distinctive roles of the prosecutor, the defence counsel and the judge, while the questioning techniques used by the justification reviewer are inspired by lawyers' examination techniques in court room and argumentation schemes. Experiments are performed on toppled poles from road scene images in the Urban Issue dataset, and wider general adoption through subsets of the PhD dataset annotated on COCO2014. Experiments show that our solution achieved a superior overturning rate of 30% and 2.3 percentage points increase in F1 score in the domain-specific application, with 50% less time required than the closest multi-agent solution. Comparable detection performance and efficient resource consumption were also seen in the general adoption. vision language model multi-agents hallucination argumentation Full Text Additional Declarations No competing interests reported. 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-8949950","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":614185742,"identity":"d0c7333a-7f0f-42c3-a71d-69ae0b98bcb6","order_by":0,"name":"Percy Lam","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuElEQVRIiWNgGAWjYDACCebGxz8q4NwEYrQwNhsznCFRS5s0YxspWsylGxukC+fZyes2MD/8wNiWRliL5ZyDDcYztyUbbjvAZgy0MYewFoMbiQ0JvNsOMG47wGDGwNhWQZyWA7xzDthvO8D+jWgtjc28DQcStx3gAdlChMMsZyQ2M844lpy87TBPsUTCOSK8by6RfPzHhxo7223H2zd++FCWTITD4CxmBuIiEknLKBgFo2AUjAJcAABDDzsiEyKQswAAAABJRU5ErkJggg==","orcid":"","institution":"University of Cambridge","correspondingAuthor":true,"prefix":"","firstName":"Percy","middleName":"","lastName":"Lam","suffix":""},{"id":614185744,"identity":"7871c084-366f-47b3-a7e7-ec6043e63510","order_by":1,"name":"Lavindra de Silva","email":"","orcid":"","institution":"University of Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Lavindra","middleName":"","lastName":"de Silva","suffix":""},{"id":614185746,"identity":"9c28fc68-c8eb-43f0-a5cc-beecbefacd0b","order_by":2,"name":"Weiwei Chen","email":"","orcid":"","institution":"University College London","correspondingAuthor":false,"prefix":"","firstName":"Weiwei","middleName":"","lastName":"Chen","suffix":""},{"id":614185752,"identity":"86890aae-38e7-4a37-a1a2-67d59b3e2b2e","order_by":3,"name":"Ioannis Brilakis","email":"","orcid":"","institution":"University of Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Ioannis","middleName":"","lastName":"Brilakis","suffix":""}],"badges":[],"createdAt":"2026-02-23 18:38:54","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8949950/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8949950/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106414838,"identity":"e8d70cc1-4a40-45d8-a985-b6546e58ff63","added_by":"auto","created_at":"2026-04-08 10:28:11","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1030233,"visible":true,"origin":"","legend":"","description":"","filename":"PercyJAAMAS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8949950/v1_covered_b10e0c16-48a0-4cdc-90f4-135ef72a2b41.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Multi-agent Court to Mitigate VLM Hallucinations","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":"vision language model, multi-agents, hallucination, argumentation","lastPublishedDoi":"10.21203/rs.3.rs-8949950/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8949950/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Hallucinations have hindered the widespread use of vision language models (VLMs) for domain-specific applications such as road maintenance. While previous researchers constructed multiple solutions for different sources of visual hallucinations, knowledge gaps persist in handling context-dependent hallucinations where the targeted objects are difficult to be prompted precisely. This research explores hallucination by converting image-to-text binary classifications into evidential arguments by VLM agents, each providing a binary Yes/No answer with a justification. The proposed solution involves VLM agents performing distinct roles, starting with a detection unit that uses a primary detector and a reviewer to verify scope compatibility. These agents interact to aggregate their findings and justifications into a single, unified verdict. The different roles of each agent are inspired by the distinctive roles of the prosecutor, the defence counsel and the judge, while the questioning techniques used by the justification reviewer are inspired by lawyers' examination techniques in court room and argumentation schemes. Experiments are performed on toppled poles from road scene images in the Urban Issue dataset, and wider general adoption through subsets of the PhD dataset annotated on COCO2014. Experiments show that our solution achieved a superior overturning rate of 30\\% and 2.3 percentage points increase in F1 score in the domain-specific application, with 50\\% less time required than the closest multi-agent solution. Comparable detection performance and efficient resource consumption were also seen in the general adoption.","manuscriptTitle":"A Multi-agent Court to Mitigate VLM Hallucinations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-31 03:25:39","doi":"10.21203/rs.3.rs-8949950/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":"aa1e087d-a823-4c2e-8ca0-09dd96b82088","owner":[],"postedDate":"March 31st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-01T15:24:41+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-31 03:25:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8949950","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8949950","identity":"rs-8949950","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}