Program-Guided Refinement with Debate: A Multi-Agent LLM-Based Automated Fact-Checking Model

Program-Guided Refinement with Debate: A Multi-Agent LLM-Based Automated Fact-Checking 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 Research Article Program-Guided Refinement with Debate: A Multi-Agent LLM-Based Automated Fact-Checking Model Tao Xue, Wenzhuo Liu, Long Xi, Wen Lv This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8033646/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 11 You are reading this latest preprint version Abstract The explosive spread of online information has led to the proliferation of false claims, which makes automated fact-checking increasingly urgent. Existing automated fact-checking models have three common problems: the lack of interpretability, the hallucination phenomenon and the lower inference efficiency. In this paper, we propose a novel model, PGR-Debate (Program-Guided Refinement with Debate), to address these challenges. By designing a multi-agent debate, the PGR-Debate decomposes complex claims into three executable sub-tasks: Question, Verify and Predict, thereby significantly enhancing the interpretability of fact-checking. To alleviate the hallucination problem, we design two Debater agents and one Finalizer agent. The two Debater agents engage in interactive debates to identify and correct errors in the reasoning program. The Finalizer then rewrites the program, gradually improving the faithfulness and credibility of explanations. To accelerate inference and enable lightweight deployment, we adopt a knowledge distillation strategy. A high-performance model serves as the teacher, and a task-aware distillation framework transfers its multi-hop reasoning capability to a smaller student model. This approach improves inference efficiency while preserving reasoning consistency. The model requires neither domain-specific pretraining nor task-specific fine-tuning, but leverages instruction-based prompting and knowledge distillation. Experiments on the standard FEVEROUS-S and HOVER datasets demonstrate that PGR-Debate outperforms multiple baselines under different evidence availability settings (Gold, Open-book), reduces reasoning time to 30%–50% of traditional methods, and boosts the student model’s inference speed by 1.9× after distillation. Moreover, the error rate of predictions is reduced by about 50%, with the semantic error rate dropping from 6.8% to 2.88% after distillation. Experimental results on the HOVER dataset demonstrate that, compared with the ProgramFC baseline (Qwen2.5-14B), our PGR-Debate model improves explanation faithfulness by 42–43 percentage points (about 3.3×) at the sentence level and by 7–8 percentage points (about 1.4×) at the program level, significantly enhancing the factual consistency of reasoning chains. Fact checking LLM Multi-agent debate Claim decomposition Distillation Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 18 May, 2026 Reviews received at journal 27 Apr, 2026 Reviews received at journal 25 Apr, 2026 Reviewers agreed at journal 25 Apr, 2026 Reviewers agreed at journal 22 Apr, 2026 Reviews received at journal 28 Dec, 2025 Reviewers agreed at journal 16 Dec, 2025 Reviewers invited by journal 16 Dec, 2025 Editor assigned by journal 09 Nov, 2025 Submission checks completed at journal 05 Nov, 2025 First submitted to journal 04 Nov, 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-8033646","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":562633572,"identity":"8bd71e20-491b-4cfb-b2a4-cbaced298e3a","order_by":0,"name":"Tao Xue","email":"","orcid":"","institution":"Xi’an Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Xue","suffix":""},{"id":562633574,"identity":"7efb81e2-8f03-44dd-967a-ffe26738334f","order_by":1,"name":"Wenzhuo Liu","email":"","orcid":"","institution":"Xi’an Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Wenzhuo","middleName":"","lastName":"Liu","suffix":""},{"id":562633575,"identity":"8b9208ae-e88d-4a0d-9fb4-4fac944d68fa","order_by":2,"name":"Long Xi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAqElEQVRIiWNgGAWjYBACPgbGxgcQZgKRWtgYGJsNSNXCwCZBohaJ5LZq3h2HGfjZcwwYfu4gRgvPwbabM88cZpDseWPA2HuGGC3sjW03PrYdZjC4kWPAzNhGjBagsoJEoBZ74rUAbWEA2yJBtBaeg82SM8+k80iceVZwsJcYLfwS6Q8/8+6wluNvT9744CcxWsCAsYGBB0QfIFYDWMsoGAWjYBSMAtwAAMGOMPTJ5bcsAAAAAElFTkSuQmCC","orcid":"","institution":"Xi’an Polytechnic University","correspondingAuthor":true,"prefix":"","firstName":"Long","middleName":"","lastName":"Xi","suffix":""},{"id":562633576,"identity":"75479d0c-702d-47fd-9748-9f77359d2acf","order_by":3,"name":"Wen Lv","email":"","orcid":"","institution":"Xi’an Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Wen","middleName":"","lastName":"Lv","suffix":""}],"badges":[],"createdAt":"2025-11-05 03:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8033646/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8033646/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98628613,"identity":"b1de2de6-4e38-4d7d-9a4a-4b83677c7b28","added_by":"auto","created_at":"2025-12-19 17:11:52","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6719,"visible":true,"origin":"","legend":"","description":"","filename":"4d1f5910076f4294b78e876b28bff918.json","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/4bb13989e7913569e0a1b4be.json"},{"id":98615431,"identity":"8027f2a4-4f4c-416d-8881-67849bd06827","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136682,"visible":true,"origin":"","legend":"","description":"","filename":"4d1f5910076f4294b78e876b28bff9181enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/2e86684a81191c39f46c8adf.xml"},{"id":98615429,"identity":"d057c7bb-3812-40fa-a567-18b51677125b","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19634,"visible":true,"origin":"","legend":"","description":"","filename":"AblationDebateTurn.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/655ca85b3daebd2ee5b0bee8.png"},{"id":98615433,"identity":"9f4e3e2c-8ff0-4ec1-9723-2c910858781f","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15426,"visible":true,"origin":"","legend":"","description":"","filename":"AblationRefinement.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/0299231cf882444c19eea63f.png"},{"id":98629269,"identity":"6c29285e-9a09-4d1f-89e1-cfdb2190ffda","added_by":"auto","created_at":"2025-12-19 17:13:28","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1376681,"visible":true,"origin":"","legend":"","description":"","filename":"AllFreamworks.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/fc0927daed908668ed20827f.png"},{"id":98615435,"identity":"63d90b7e-a9e8-4b31-a2b2-fc8c8f8befb2","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":748648,"visible":true,"origin":"","legend":"","description":"","filename":"ClaimDecompose.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/b9a09438b5e7091fc8aa00a6.png"},{"id":98629109,"identity":"b2e5527f-9487-4e46-af1c-a6b36a3adbe4","added_by":"auto","created_at":"2025-12-19 17:13:15","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40405,"visible":true,"origin":"","legend":"","description":"","filename":"Coverletter.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/a93dea64b597ef7b80d0856a.pdf"},{"id":98615437,"identity":"2bcf45be-c24e-47e0-9cbf-60168521b443","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":844040,"visible":true,"origin":"","legend":"","description":"","filename":"DebateandRefine.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/378ca57874c82d31d2df990d.png"},{"id":98629625,"identity":"03a60da4-a387-4978-970d-a9e074d65612","added_by":"auto","created_at":"2025-12-19 17:14:21","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39950,"visible":true,"origin":"","legend":"","description":"","filename":"PGRDebateFEVEROUSConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/32ae5b785c17ffe91b4fe0ac.png"},{"id":98615436,"identity":"043605e0-32a0-4045-a572-f93499785780","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38971,"visible":true,"origin":"","legend":"","description":"","filename":"PGRDebateHOVERConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/544d4dc8b707cde32cec1674.png"},{"id":98629298,"identity":"7838415c-cb93-4f6c-b3c0-9ee71e5843aa","added_by":"auto","created_at":"2025-12-19 17:13:33","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":553047,"visible":true,"origin":"","legend":"","description":"","filename":"ProgramFCExample.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/d06aad107312a2e2591ba7e0.png"},{"id":98615439,"identity":"e5d557c4-b0a0-401f-98de-80e27f696a16","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39298,"visible":true,"origin":"","legend":"","description":"","filename":"ProgramFCFEVEROUSConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/8a89aaef7c8a1bfb33cac29c.png"},{"id":98629202,"identity":"b0ca0790-cfbb-4a97-add6-e8937daf8f07","added_by":"auto","created_at":"2025-12-19 17:13:25","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37166,"visible":true,"origin":"","legend":"","description":"","filename":"ProgramFCHOVERConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/2446fbc4e44202cff9df41fc.png"},{"id":98629508,"identity":"8eb0b7e9-9ff2-4f60-a98e-a0c32fe95ee8","added_by":"auto","created_at":"2025-12-19 17:14:04","extension":"pdf","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2692668,"visible":true,"origin":"","legend":"","description":"","filename":"ProgramGuidedRefinementwithDebateAMultiAgentLLMBasedAutomatedFactCheckingModel.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/da4eae84773ed16d49efc7c6.pdf"},{"id":98615440,"identity":"85bdbe15-1f83-47ae-8596-67cf7b5078d5","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"eps","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2890,"visible":true,"origin":"","legend":"","description":"","filename":"empty.eps","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/8d32069b52e819252bbb3d50.eps"},{"id":98628422,"identity":"396e0529-cbfc-4475-b2b3-7f0d197aeb32","added_by":"auto","created_at":"2025-12-19 17:11:31","extension":"eps","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91593,"visible":true,"origin":"","legend":"","description":"","filename":"fig.eps","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/913b7c78bf605b6cf81ed46a.eps"},{"id":98628881,"identity":"0977f08e-b365-4782-838e-3d4dde4f143c","added_by":"auto","created_at":"2025-12-19 17:12:43","extension":"bst","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":146013,"visible":true,"origin":"","legend":"","description":"","filename":"snapacite.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/7ac086c1c8db6ad735ee648c.bst"},{"id":98628822,"identity":"191e7a96-11f4-4786-8f65-f8bea5e3fb31","added_by":"auto","created_at":"2025-12-19 17:12:34","extension":"bst","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29828,"visible":true,"origin":"","legend":"","description":"","filename":"snaps.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/38f34c59bd18ba29250bfc45.bst"},{"id":98615451,"identity":"65cbbc7f-e7d8-4997-a710-b302b45c0968","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"pdf","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":421391,"visible":true,"origin":"","legend":"","description":"","filename":"snarticle.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/f0507257143b6ffda22c0761.pdf"},{"id":98629681,"identity":"c0c775fc-44bd-465c-afb9-36688ad3c81e","added_by":"auto","created_at":"2025-12-19 17:14:29","extension":"bst","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35515,"visible":true,"origin":"","legend":"","description":"","filename":"snbasic.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/f30cae96ab47c5657ec8ced3.bst"},{"id":98615443,"identity":"2f36fc67-7713-4858-a137-8f1ff4978478","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"bst","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33968,"visible":true,"origin":"","legend":"","description":"","filename":"snchicago.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/773916f6a607c8227ca392cd.bst"},{"id":98615452,"identity":"7d32d7e4-539c-4e31-9f4a-786c805ebe8c","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"cls","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55857,"visible":true,"origin":"","legend":"","description":"","filename":"snjnl.cls","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/ff40d747109e8ef0acb29419.cls"},{"id":98615444,"identity":"d90474e4-b5b9-4637-b400-19a9fb50abee","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"bst","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64023,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysay.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/7d602031d670ae0224ed6b48.bst"},{"id":98615447,"identity":"860d1797-b722-449d-9e37-5a3109fc849f","added_by":"auto","created_at":"2025-12-19 15:11:57","extension":"bst","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64166,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysnum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/3e9fd3013b970f5c44a57c90.bst"},{"id":98629629,"identity":"8c1a6216-6f24-470f-a557-eb1afeaa5823","added_by":"auto","created_at":"2025-12-19 17:14:21","extension":"bst","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37333,"visible":true,"origin":"","legend":"","description":"","filename":"snnature.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/c41f7080f5612e8d81c07ec0.bst"},{"id":98628676,"identity":"8ea52ee4-acf7-4642-84f3-e3bbe3fe8aba","added_by":"auto","created_at":"2025-12-19 17:12:02","extension":"bst","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39951,"visible":true,"origin":"","legend":"","description":"","filename":"snvancouveray.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/d053735b39d973c2bf6d37e8.bst"},{"id":98629643,"identity":"0094b6cd-0ef0-4938-a72c-2e84b4e9be32","added_by":"auto","created_at":"2025-12-19 17:14:24","extension":"bst","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40758,"visible":true,"origin":"","legend":"","description":"","filename":"snvancouvernum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/4e1a92c2dd0d2fda2f788932.bst"},{"id":98615459,"identity":"eb3bfb60-c169-4c65-ac22-4aa7615233ab","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"pdf","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":418495,"visible":true,"origin":"","legend":"","description":"","filename":"usermanual.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/0b924af4ce7130dcdedecdf3.pdf"},{"id":98615461,"identity":"bb6a6a91-fd6c-4569-919a-4189a30cf061","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34817,"visible":true,"origin":"","legend":"","description":"","filename":"OnlinePGRDebateFEVEROUSConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/47776d3f2d73bf3df4d6a79f.png"},{"id":98628912,"identity":"63cf7eb5-8e93-484c-8307-3f5b6a167ca5","added_by":"auto","created_at":"2025-12-19 17:12:47","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33976,"visible":true,"origin":"","legend":"","description":"","filename":"OnlinePGRDebateHOVERConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/2a9cda4b3b2d42d0e0c90092.png"},{"id":98615458,"identity":"afd58d79-01d1-4124-ad81-12659eb174cf","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34351,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineProgramFCFEVEROUSConfusionMatrix.png","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/8caa0ffff8d10f582c16959c.png"},{"id":98615455,"identity":"1673899d-9099-4675-860e-5546e561496e","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"xml","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":176269,"visible":true,"origin":"","legend":"","description":"","filename":"4d1f5910076f4294b78e876b28bff9181structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/eebc97427178d33e77e250ce.xml"},{"id":98615460,"identity":"eebbdfb9-b54e-42f9-b753-269a7bccca58","added_by":"auto","created_at":"2025-12-19 15:11:58","extension":"html","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":158232,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1/423674aace2613a5c4979ba0.html"},{"id":98775578,"identity":"03d4131f-7cea-44ed-8c52-a285d27359d7","added_by":"auto","created_at":"2025-12-22 12:20:22","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2978408,"visible":true,"origin":"","legend":"","description":"","filename":"ProgramGuidedRefinementwithDebateAMultiAgentLLMBasedAutomatedFactCheckingModel.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033646/v1_covered_738279c4-87e5-4a75-8f5a-97d550cd1fb7.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Program-Guided Refinement with Debate: A Multi-Agent LLM-Based Automated Fact-Checking 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":"international-journal-of-machine-learning-and-cybernetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmlc","sideBox":"Learn more about [International Journal of Machine Learning and Cybernetics](http://actavetscand.biomedcentral.com/)","snPcode":"13042","submissionUrl":"https://submission.nature.com/new-submission/13042/3","title":"International Journal of Machine Learning and Cybernetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Fact checking, LLM, Multi-agent debate, Claim decomposition, Distillation","lastPublishedDoi":"10.21203/rs.3.rs-8033646/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8033646/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe explosive spread of online information has led to the proliferation of false claims, which makes automated fact-checking increasingly urgent. Existing automated fact-checking models have three common problems: the lack of interpretability, the hallucination phenomenon and the lower inference efficiency. In this paper, we propose a novel model, PGR-Debate (Program-Guided Refinement with Debate), to address these challenges. By designing a multi-agent debate, the PGR-Debate decomposes complex claims into three executable sub-tasks: Question, Verify and Predict, thereby significantly enhancing the interpretability of fact-checking. To alleviate the hallucination problem, we design two Debater agents and one Finalizer agent. The two Debater agents engage in interactive debates to identify and correct errors in the reasoning program. The Finalizer then rewrites the program, gradually improving the faithfulness and credibility of explanations. To accelerate inference and enable lightweight deployment, we adopt a knowledge distillation strategy. A high-performance model serves as the teacher, and a task-aware distillation framework transfers its multi-hop reasoning capability to a smaller student model. This approach improves inference efficiency while preserving reasoning consistency. The model requires neither domain-specific pretraining nor task-specific fine-tuning, but leverages instruction-based prompting and knowledge distillation. Experiments on the standard FEVEROUS-S and HOVER datasets demonstrate that PGR-Debate outperforms multiple baselines under different evidence availability settings (Gold, Open-book), reduces reasoning time to 30%\u0026ndash;50% of traditional methods, and boosts the student model\u0026rsquo;s inference speed by 1.9\u0026times; after distillation. Moreover, the error rate of predictions is reduced by about 50%, with the semantic error rate dropping from 6.8% to 2.88% after distillation. Experimental results on the HOVER dataset demonstrate that, compared with the ProgramFC baseline (Qwen2.5-14B), our PGR-Debate model improves explanation faithfulness by 42\u0026ndash;43 percentage points (about 3.3\u0026times;) at the sentence level and by 7\u0026ndash;8 percentage points (about 1.4\u0026times;) at the program level, significantly enhancing the factual consistency of reasoning chains.\u003c/p\u003e","manuscriptTitle":"Program-Guided Refinement with Debate: A Multi-Agent LLM-Based Automated Fact-Checking Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-19 15:11:50","doi":"10.21203/rs.3.rs-8033646/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-19T01:29:40+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-27T08:21:43+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-25T13:44:02+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"174122274620773306706141439878065280736","date":"2026-04-25T13:09:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"32406639289163186621949886905919523711","date":"2026-04-23T01:21:45+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-29T00:55:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"305766987312724715183608778301964910435","date":"2025-12-17T01:40:36+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-16T13:20:37+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-09T14:09:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-11-05T06:56:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Machine Learning and Cybernetics","date":"2025-11-05T03:17:26+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-machine-learning-and-cybernetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmlc","sideBox":"Learn more about [International Journal of Machine Learning and Cybernetics](http://actavetscand.biomedcentral.com/)","snPcode":"13042","submissionUrl":"https://submission.nature.com/new-submission/13042/3","title":"International Journal of Machine Learning and Cybernetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1ed3e909-cd61-4604-b7ce-f9227c73029f","owner":[],"postedDate":"December 19th, 2025","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-19T01:29:40+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-19T01:39:42+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-19 15:11:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8033646","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8033646","identity":"rs-8033646","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}