A Comparative Analysis of Chain-of-Thought Distillation from Gemini 3 to Legacy (Flan-T5) and Modern (Gemma) SLMs for Domain-Specific Classification

A Comparative Analysis of Chain-of-Thought Distillation from Gemini 3 to Legacy (Flan-T5) and Modern (Gemma) SLMs for Domain-Specific Classification | 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 Comparative Analysis of Chain-of-Thought Distillation from Gemini 3 to Legacy (Flan-T5) and Modern (Gemma) SLMs for Domain-Specific Classification Vignesh Chinthakuntla, Sankar Ganesh Paramasivam, Tejaswini Neelarapu, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8798957/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) such as Gemini 3 demonstrate strong multi-step reasoning, but the associated memory footprint and inference latency limit suitability for real‑time, edge‑deployed financial services. Small Language Models (SLMs) enable lower-cost deployment, yet standard supervised fine‑tuning frequently fails to capture fine‑grained intent boundaries in customer support taxonomies. A comparative analysis is conducted for distillation-by-synthesis that transfers chain‑of‑thought (CoT) supervision from a Teacher LLM (Gemini 3) into two Student architectures: a legacy encoder–decoder model (Flan‑T5 Base, 250M parameters) and a modern decoder‑only model (Gemma 2B). A reasoning‑augmented training set is synthesized on Banking77 by prompting the Teacher to produce intent labels together with short, structured justifications that highlight discriminative cues (for example, separating card_arrival from card_delivery_estimate). Student models are fine‑tuned to generate both an intent label and an aligned rationale. Evaluation covers three dimensions: (1) intent accuracy, (2) reasoning fidelity measured through rubric‑based label–rationale consistency, and (3) inference latency under batch‑1 serving. Results indicate that Gemma 2B yields the strongest accuracy and the most nuanced explanations, while Flan‑T5 Base delivers a favorable deployment trade‑off by maintaining competitive accuracy with substantially lower memory demand and latency. The analysis clarifies how architectural bias (encoder–decoder stability versus decoder‑only generation capacity) interacts with CoT distillation, providing guidance for low‑latency intent classifiers in compliance‑sensitive banking environments. Knowledge distillation chain-of-thought small language models intent classification Banking77 edge inference explainable NLP 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-8798957","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":587412636,"identity":"194ffe1f-3a5c-43cf-be0f-07632bd6f2ea","order_by":0,"name":"Vignesh Chinthakuntla","email":"data:image/png;base64,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","orcid":"","institution":"FISAT","correspondingAuthor":true,"prefix":"","firstName":"Vignesh","middleName":"","lastName":"Chinthakuntla","suffix":""},{"id":587412637,"identity":"0208fb30-1624-481a-8f7e-93cb072061bd","order_by":1,"name":"Sankar Ganesh Paramasivam","email":"","orcid":"","institution":"Illinois Institute Of Technology","correspondingAuthor":false,"prefix":"","firstName":"Sankar","middleName":"Ganesh","lastName":"Paramasivam","suffix":""},{"id":587412638,"identity":"bf97b4d5-d287-4afa-aa54-d21f784fb7c1","order_by":2,"name":"Tejaswini Neelarapu","email":"","orcid":"","institution":"Illinois Institute Of Technology","correspondingAuthor":false,"prefix":"","firstName":"Tejaswini","middleName":"","lastName":"Neelarapu","suffix":""},{"id":587412639,"identity":"a8570432-38ae-4049-8e91-c9112f6c2ce8","order_by":3,"name":"Jagadesh Radhakrishnan","email":"","orcid":"","institution":"SRM Institute of technology","correspondingAuthor":false,"prefix":"","firstName":"Jagadesh","middleName":"","lastName":"Radhakrishnan","suffix":""}],"badges":[],"createdAt":"2026-02-05 15:39:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8798957/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8798957/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103506238,"identity":"49499f4e-b877-4ce5-b36b-c9fea7d2f24a","added_by":"auto","created_at":"2026-02-26 13:34:48","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":373106,"visible":true,"origin":"","legend":"","description":"","filename":"CoTDistillationBanking77Paperv2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8798957/v1_covered_a77598d8-8b1d-4aa9-b764-89dc4e16ed47.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Comparative Analysis of Chain-of-Thought Distillation from Gemini 3 to Legacy (Flan-T5) and Modern (Gemma) SLMs for Domain-Specific Classification","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":"Knowledge distillation, chain-of-thought, small language models, intent classification, Banking77, edge inference, explainable NLP","lastPublishedDoi":"10.21203/rs.3.rs-8798957/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8798957/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLarge Language Models (LLMs) such as Gemini 3 demonstrate strong multi-step reasoning, but the associated memory footprint and inference latency limit suitability for real‑time, edge‑deployed financial services. Small Language Models (SLMs) enable lower-cost deployment, yet standard supervised fine‑tuning frequently fails to capture fine‑grained intent boundaries in customer support taxonomies. A comparative analysis is conducted for distillation-by-synthesis that transfers chain‑of‑thought (CoT) supervision from a Teacher LLM (Gemini 3) into two Student architectures: a legacy encoder\u0026ndash;decoder model (Flan‑T5 Base, 250M parameters) and a modern decoder‑only model (Gemma 2B).\u003c/p\u003e \u003cp\u003eA reasoning‑augmented training set is synthesized on Banking77 by prompting the Teacher to produce intent labels together with short, structured justifications that highlight discriminative cues (for example, separating card_arrival from card_delivery_estimate). Student models are fine‑tuned to generate both an intent label and an aligned rationale. Evaluation covers three dimensions: (1) intent accuracy, (2) reasoning fidelity measured through rubric‑based label\u0026ndash;rationale consistency, and (3) inference latency under batch‑1 serving. Results indicate that Gemma 2B yields the strongest accuracy and the most nuanced explanations, while Flan‑T5 Base delivers a favorable deployment trade‑off by maintaining competitive accuracy with substantially lower memory demand and latency. The analysis clarifies how architectural bias (encoder\u0026ndash;decoder stability versus decoder‑only generation capacity) interacts with CoT distillation, providing guidance for low‑latency intent classifiers in compliance‑sensitive banking environments.\u003c/p\u003e","manuscriptTitle":"A Comparative Analysis of Chain-of-Thought Distillation from Gemini 3 to Legacy (Flan-T5) and Modern (Gemma) SLMs for Domain-Specific Classification","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-10 09:13:48","doi":"10.21203/rs.3.rs-8798957/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":"9d965786-bfa7-417d-9ab9-6dea41107281","owner":[],"postedDate":"February 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-03T15:22:15+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-10 09:13:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8798957","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8798957","identity":"rs-8798957","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}