DRaFT-Q: Dynamic Rank-Aware Fine-Tuning under Quantization for Efficient and Reward-Sensitive Adaptation of Language Models

DRaFT-Q: Dynamic Rank-Aware Fine-Tuning under Quantization for Efficient and Reward-Sensitive Adaptation of Language Models | 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 DRaFT-Q: Dynamic Rank-Aware Fine-Tuning under Quantization for Efficient and Reward-Sensitive Adaptation of Language Models Adharapurapu V S M Ashok Kumar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7491496/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 Large Language Models (LLMs) demonstrate remarkable capabilities in language understanding, reasoning, and cross-domain generalization. However, their immense scale makes full fine-tuning computationally and memory-intensive. Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA and QLoRA address this by introducing low-rank adapters while keeping most of the model frozen. Yet, these methods typically use fixed adapter ranks, which do not adapt to varying layer complexity or evolving training dynamics—resulting in suboptimal convergence and inefficient capacity utilization. We introduce DRaFT-Q (Dynamic Rank-Aware Fine-Tuning under Quantization), a novel PEFT strategy that combines dynamic LoRA rank allocation with token-level reward-weighted loss. DRaFTQ adjusts adapter rank in real-time based on lightweight curvature signals from gradient statistics and prioritizes semantically important tokens during learning using external or task-derived reward weights. This dual adaptation improves both parameter usage and training focus under strict memory constraints. We evaluate DRaFT-Q on reasoning and QA benchmarks including CommonsenseQA, OpenBookQA, GSM8K, OpenAssistant, and OpenHermes using LLaMA-2-7B and 13B models under 4-bit quantization. Experiments on constrained GPUs (T4 for 7B; L40s for 13B) show DRaFT-Q achieves better loss convergence, improved generalization, and competitive accuracy compared to LoRA, QLoRA, and AdaLoRA—while maintaining comparable resource usage. Our findings highlight DRaFT-Q’s effectiveness for dynamic, reward-aware fine-tuning of quantized LLMs at scale. Artificial Intelligence and Machine Learning Dynamic Rank Allocation Quantization Parameter-Efficient Fine-Tuning LoRA QLoRA Reward-Weighted Learning Adaptive PEFT Large Language Models CommonsenseQA OpenBookQA GSM8K Full Text Additional Declarations The authors declare no competing interests. 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-7491496","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":507625177,"identity":"b288351f-2d31-4401-b842-3a8e3e984643","order_by":0,"name":"Adharapurapu V S M Ashok Kumar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+UlEQVRIiWNgGAWjYPCCAwwSPGCGDRAzNh4gRUsaSEsDSVoOQ7l4AP+M9Kcbfvy5IyfZc8Z0w8895+3Wth8G2lJjE41Li8SNHLObvW3PjKV5e8xu9jy7nbztTCJQy7G03AZces6cYbvB23A4cR4/j9kNngO3k80OALUwNhzGqUX+zPFnN//8gWi5+efAuWSz8w/xazE43mB2m4ftcOJsoMNu8xw4YGd2g4AthseBKmXbDhtL9hwruy1zIDnB7AbQlgQ8fpE7zP7s5ps/h+UkziRvu/nmgJ292fn0hw8+1Njg9j46SASrTCBWOQjYk6J4FIyCUTAKRgYAAPV/bnY9xI4qAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0008-7946-2908","institution":"Koneru Lakshmaiah University","correspondingAuthor":true,"prefix":"","firstName":"Adharapurapu","middleName":"V S M Ashok","lastName":"Kumar","suffix":""}],"badges":[],"createdAt":"2025-08-29 20:56:53","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":true,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-7491496/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7491496/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90476835,"identity":"86e2f44e-7031-4a0f-9cc1-03d9c5842da1","added_by":"auto","created_at":"2025-09-03 07:11:54","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":334251,"visible":true,"origin":"","legend":"","description":"","filename":"DRaFTQ.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7491496/v1_covered_08cf20ee-4147-49cf-a07a-7ad15b92cc2e.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eDRaFT-Q: Dynamic Rank-Aware Fine-Tuning under Quantization for Efficient and Reward-Sensitive Adaptation of Language Models\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Koneru Lakshmaiah University","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":"Dynamic Rank Allocation, Quantization, Parameter-Efficient Fine-Tuning, LoRA, QLoRA, Reward-Weighted Learning, Adaptive PEFT, Large Language Models, CommonsenseQA, OpenBookQA, GSM8K","lastPublishedDoi":"10.21203/rs.3.rs-7491496/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7491496/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLarge Language Models (LLMs) demonstrate remarkable capabilities in language understanding, reasoning, and cross-domain generalization. However, their immense scale makes full fine-tuning computationally and memory-intensive. Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA and QLoRA address this by introducing low-rank adapters while keeping most of the model frozen. Yet, these methods typically use fixed adapter ranks, which do not adapt to varying layer complexity or evolving training dynamics—resulting in suboptimal convergence and inefficient capacity utilization. We introduce DRaFT-Q (Dynamic Rank-Aware Fine-Tuning under Quantization), a novel PEFT strategy that combines dynamic LoRA rank allocation with token-level reward-weighted loss. DRaFTQ adjusts adapter rank in real-time based on lightweight curvature signals from gradient statistics and prioritizes semantically important tokens during learning using external or task-derived reward weights. This dual adaptation improves both parameter usage and training focus under strict memory constraints. We evaluate DRaFT-Q on reasoning and QA benchmarks including CommonsenseQA, OpenBookQA, GSM8K, OpenAssistant, and OpenHermes using LLaMA-2-7B and 13B models under 4-bit quantization. Experiments on constrained GPUs (T4 for 7B; L40s for 13B) show DRaFT-Q achieves better loss convergence, improved generalization, and competitive accuracy compared to LoRA, QLoRA, and AdaLoRA—while maintaining comparable resource usage. Our findings highlight DRaFT-Q’s effectiveness for dynamic, reward-aware fine-tuning of quantized LLMs at scale.\u003c/p\u003e","manuscriptTitle":"DRaFT-Q: Dynamic Rank-Aware Fine-Tuning under Quantization for Efficient and Reward-Sensitive Adaptation of Language Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-03 07:03:46","doi":"10.21203/rs.3.rs-7491496/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":"8e9adc18-9ecd-4c9e-ba9e-b44faf80257d","owner":[],"postedDate":"September 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":53924541,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-09-03T07:03:46+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-03 07:03:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7491496","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7491496","identity":"rs-7491496","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}