LegalMALR:Multi-Agent Query Understanding and LLM-Based Reranking for Chinese Statute Retrieval

LegalMALR:Multi-Agent Query Understanding and LLM-Based Reranking for Chinese Statute Retrieval | 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 LegalMALR:Multi-Agent Query Understanding and LLM-Based Reranking for Chinese Statute Retrieval Yunhan Li, Mingjie Xie, Gaoli Kang, Zihan Gong, Gengshen Wu, Min Yang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8425174/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 Statute retrieval is essential for legal assistance and judicial decision support, yet real-world legal queries are often implicit, multi-issue, and expressed in colloquial or underspecified forms. These characteristics make it difficult for conventional retrieval-augmented generation pipelines to recover the statutory elements required for accurate retrieval. Dense retrievers focus primarily on the literal surface form of the query, whereas lightweight rerankers lack the legal-reasoning capacity needed to assess statutory applicability. We present LegalMALR, a retrieval framework that integrates a Multi-Agent Query Understanding System (MAS) with a zero-shot large-language-model-based reranking module (LLM Reranker). MAS generates diverse, legally grounded reformulations and conducts iterative dense retrieval to broaden candidate coverage. To stabilise the stochastic behaviour of LLM-generated rewrites, we optimise a unified MAS policy using Generalized Reinforcement Policy Optimization(GRPO). The accumulated candidate set is subsequently evaluated by the LLM Reranker, which performs natural-language legal reasoning to produce the final ranking. We further construct CSAID, a dataset of 118 difficult Chinese legal queries annotated with multiple statutory labels, and evaluate LegalMALR on both CSAID and the public STARD benchmark. Experiments show that LegalMALR substantially outperforms strong Retrieval-augmented generation(RAG) baselines in both in-distribution and out-of-distribution settings, demonstrating the effectiveness of combining multi-perspective query interpretation, reinforcement-based policy optimisation, and large-model reranking for statute retrieval. Chinese statute retrieval multi-agent query understanding GRPO LLM-based reranking Chinese statute dataset 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-8425174","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":577640578,"identity":"104d3ed8-d8cd-4762-bec3-2bb99aeaccdc","order_by":0,"name":"Yunhan Li","email":"","orcid":"","institution":"City University of Macau","correspondingAuthor":false,"prefix":"","firstName":"Yunhan","middleName":"","lastName":"Li","suffix":""},{"id":577640579,"identity":"f0810f58-7711-4f80-ac47-f31ac26b0e6e","order_by":1,"name":"Mingjie Xie","email":"","orcid":"","institution":"Southern University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Mingjie","middleName":"","lastName":"Xie","suffix":""},{"id":577640580,"identity":"06bedcad-962b-4d7b-86f9-bec0f3d92d11","order_by":2,"name":"Gaoli Kang","email":"","orcid":"","institution":"Shenzhen University of Advanced Technology","correspondingAuthor":false,"prefix":"","firstName":"Gaoli","middleName":"","lastName":"Kang","suffix":""},{"id":577640581,"identity":"fe6ca217-316d-4b92-a861-0e93ae8c5e2f","order_by":3,"name":"Zihan Gong","email":"","orcid":"","institution":"Shenzhen Institutes of Advanced Technology","correspondingAuthor":false,"prefix":"","firstName":"Zihan","middleName":"","lastName":"Gong","suffix":""},{"id":577640582,"identity":"311f2edd-352a-428b-8287-7008abc487de","order_by":4,"name":"Gengshen Wu","email":"","orcid":"","institution":"City University of Macau","correspondingAuthor":false,"prefix":"","firstName":"Gengshen","middleName":"","lastName":"Wu","suffix":""},{"id":577640583,"identity":"bec7701a-f25e-418b-bf95-63382c8d99df","order_by":5,"name":"Min Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1ElEQVRIiWNgGAWjYDACZijNz8DYQKIWyQaglgMk2WYAUk6UFt123gPMPDV37DafP9z4+QODnTwD+1n8Os0O8yUw8xx7lrztRmKzxAGGZMMGnrwEAlp4DJh52A4nm91gbABqYU5gkOAxIELLv8PJxv0Hm38cYKgnUgtv22E7A4bENqAth4nTcnBu3+EEiRuJbRZnDI4btvHkENBy/ozhgzffDtvz9x9/fKOiolqen/0Mfi0gcIiHgSGxAcwEKmYjqB4IGH8wMNgTo3AUjIJRMApGKAAANbBCy71//B4AAAAASUVORK5CYII=","orcid":"","institution":"Shenzhen Institutes of Advanced Technology","correspondingAuthor":true,"prefix":"","firstName":"Min","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2025-12-22 12:54:59","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8425174/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8425174/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100850065,"identity":"261c434e-9ed2-4d5d-9d75-4f85c92ccf15","added_by":"auto","created_at":"2026-01-22 05:47:19","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7968,"visible":true,"origin":"","legend":"","description":"","filename":"c6377675b1ef42d798bc2c2cfaf74b92.json","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/033ab80cba51bfb85824aed3.json"},{"id":101202509,"identity":"f5ff7524-2df4-4810-b7ba-0a540e0d1755","added_by":"auto","created_at":"2026-01-27 09:35:26","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":130565,"visible":true,"origin":"","legend":"","description":"","filename":"c6377675b1ef42d798bc2c2cfaf74b921enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/70099259e40a25f54e7b8660.xml"},{"id":100850066,"identity":"9e823479-fba1-4c97-911e-2ac92c3b1287","added_by":"auto","created_at":"2026-01-22 05:47:19","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":544064,"visible":true,"origin":"","legend":"","description":"","filename":"LegalMALRblined.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/06c2b2114aec49721e1d705f.pdf"},{"id":100850064,"identity":"b26dc2a3-b73d-4805-9c36-bb26629af685","added_by":"auto","created_at":"2026-01-22 05:47:19","extension":"cls","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55334,"visible":true,"origin":"","legend":"","description":"","filename":"snjnl.cls","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/1eb56582fb01c757f9fa001c.cls"},{"id":100850069,"identity":"4dfee573-bdfb-4b0c-9776-f3f808b41817","added_by":"auto","created_at":"2026-01-22 05:47:19","extension":"bst","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64141,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysnum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/2f746bdd68b847bcf3838ccc.bst"},{"id":100850068,"identity":"cdc7a7e8-adce-4b87-a6f5-6c1090ee436c","added_by":"auto","created_at":"2026-01-22 05:47:19","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":133940,"visible":true,"origin":"","legend":"","description":"","filename":"c6377675b1ef42d798bc2c2cfaf74b921structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/06d6b4c1cb17c42633a33ba3.xml"},{"id":100850070,"identity":"1d7845dd-9065-49b6-9218-2cae6c374b16","added_by":"auto","created_at":"2026-01-22 05:47:22","extension":"html","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":142583,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1/7e5f1809126e0aeee8656995.html"},{"id":101751131,"identity":"7e0f06f5-b571-4135-af47-78ca310ea90e","added_by":"auto","created_at":"2026-02-03 10:16:15","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":531048,"visible":true,"origin":"","legend":"","description":"","filename":"LegalMALRblined.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8425174/v1_covered_001800df-8d59-481a-b327-3b458b5d5bc9.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"LegalMALR:Multi-Agent Query Understanding and LLM-Based Reranking for Chinese Statute Retrieval","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"Chinese statute retrieval, multi-agent query understanding, GRPO, LLM-based reranking, Chinese statute dataset","lastPublishedDoi":"10.21203/rs.3.rs-8425174/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8425174/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Statute retrieval is essential for legal assistance and judicial decision support, yet real-world legal queries are often implicit, multi-issue, and expressed in colloquial or underspecified forms. These characteristics make it difficult for conventional retrieval-augmented generation pipelines to recover the statutory elements required for accurate retrieval. Dense retrievers focus primarily on the literal surface form of the query, whereas lightweight rerankers lack the legal-reasoning capacity needed to assess statutory applicability.\nWe present LegalMALR, a retrieval framework that integrates a Multi-Agent Query Understanding System (MAS) with a zero-shot large-language-model-based reranking module (LLM Reranker). MAS generates diverse, legally grounded reformulations and conducts iterative dense retrieval to broaden candidate coverage. To stabilise the stochastic behaviour of LLM-generated rewrites, we optimise a unified MAS policy using Generalized Reinforcement Policy Optimization(GRPO). The accumulated candidate set is subsequently evaluated by the LLM Reranker, which performs natural-language legal reasoning to produce the final ranking.\nWe further construct CSAID, a dataset of 118 difficult Chinese legal queries annotated with multiple statutory labels, and evaluate LegalMALR on both CSAID and the public STARD benchmark. Experiments show that LegalMALR substantially outperforms strong Retrieval-augmented generation(RAG) baselines in both in-distribution and out-of-distribution settings, demonstrating the effectiveness of combining multi-perspective query interpretation, reinforcement-based policy optimisation, and large-model reranking for statute retrieval.","manuscriptTitle":"LegalMALR:Multi-Agent Query Understanding and LLM-Based Reranking for Chinese Statute Retrieval","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-22 05:47:10","doi":"10.21203/rs.3.rs-8425174/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":"cf81422d-21e1-4656-b0cc-325fe551ca88","owner":[],"postedDate":"January 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-22T05:47:10+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-22 05:47:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8425174","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8425174","identity":"rs-8425174","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}