[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-464161-105":59,"doc-detail-464161-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","knowledge-distillation-meets-reinforcement-learning-a-cluster-driven-approach-to-image-processing","Knowledge Distillation Meets Reinforcement Learning: A Cluster-Driven Approach to Image Processing","","Knowledge distillation (KD) enables training of lightweight yet effective visual models, while reinforcement learning (RL) supports adaptive learning through environment-driven interactions. The document proposes a two-stage KDRL framework combining logit/feature distillation with RL refinement to better handle dynamic, complex data distributions. The first stage uses supervised fine-tuning, and the second stage applies confidence-based and cluster-alignment rewards with dynamic loss reliance reduction. Experiments across remote sensing, medical imaging, and cross-modal retrieval report strong accuracy, retrieval, and visual-grounding gains, improving efficiency for resource-limited deployment.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/knowledge-distillation-meets-reinforcement-learning-a-cluster-driven-approach-to-image-processing/464161/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/knowledge-distillation-meets-reinforcement-learning-a-cluster-driven-approach-to-image-processing/464161.png","ImageObject",300,407,{"name":92,"@type":93},"acoobca","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-15","2026-09-30",true,{"@type":102,"interactionType":103,"userInteractionCount":39},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does the proposed KDRL framework address?","Question",{"text":112,"@type":113},"It addresses limitations of conventional KD under domain shift and task complexity, aiming to transfer robust teacher representations to a compact student without over-regularization or bias amplification.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does the framework work in two stages?",{"text":117,"@type":113},"Stage one performs supervised fine-tuning using logit and feature-based distillation. Stage two refines the trained student with reinforcement learning using confidence-based and cluster alignment rewards while dynamically reducing reliance on task loss.",{"name":119,"@type":110,"acceptedAnswer":120},"What key innovation is introduced to improve feature learning?",{"text":121,"@type":113},"The student encoder includes auxiliary layers that evaluate and reward alignment between learned characteristics and the teacher’s cluster centers, promoting robust feature learning.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},464161,1790928442,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":39,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":56,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":143,"read_time":144,"preview_page_count":19,"preview_urls":145},3985747870733,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","Article  \nKnowledge Distillation Meets Reinforcement Learning: A Cluster-Driven Approach to Image Processing  \nTitinunt Kitrungrotsakul 1, *, Yingying Xu 1 and Preeyanuch Srichola 2,3  \nAcademic Editor: Anastasios Doulamis  \nReceived: 25 November 2025  \nRevised: 21 December 2025  \nAccepted: 26 December 2025  \nPublished: 28 December 2025  \nCopyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.  \n1 Research Center for Space Computing System, Zhejiang Lab, Hangzhou 311121, China; [cs_ying@zhejianglab.org](cs_ying@zhejianglab.org)  \n2 Kasetsart Agricultural and Agro-Industrial Product Improvement Institute, Kasetsart University, Bangkok 10900, Thailand; [aappua@ku.ac.th](aappua@ku.ac.th)  \n3 Cellulose for Future Materials and Technologies Special Research Unit, Department of Biotechnology, Faculty of Agro-Industry, Kasetsart University, Bangkok 10900, Thailand  \n* Correspondence: [titinuntkitrungrotsakul@zhejianglab.com](titinuntkitrungrotsakul@zhejianglab.com)  \nAbstract  \nKnowledge distillation (KD) enables the training of lightweight yet effective models, particularly in the visual domain. Meanwhile, reinforcement learning (RL) facilitates adaptive learning through environment-driven interactions, addressing the limitations of KD in handling dynamic and complex tasks. We propose a novel two-stage framework integrating Knowledge Distillation with Reinforcement Learning (KDRL) to enhance model adaptability to complex data distributions, such as remote sensing and medical imaging. In the first stage, supervised fine-tuning guides the student model using logit and feature-based distillation. The second stage refines the model via RL, leveraging confidence-based and cluster alignment rewards while dynamically reducing reliance on task loss. By combining the strengths of supervised knowledge distillation and reinforcement learning, KDRL provides a comprehensive approach to address the dual challenges of model efficiency and domain heterogeneity. A key innovation is the introduction of auxiliary layers within the student encoder to evaluate and reward the alignment of the characteristics with the teacher’s cluster centers, promoting robust feature learning. Our framework demonstrates superior performance and computational efficiency across diverse tasks, establishing a scalable design for efficient model training. Across remote sensing benchmarks, KDRL boosts the lightweight CLIP/ViT-B-32 student to 69.51% zero-shot accuracy on AID and 80.08% on RESISC45; achieves state-of-the-art cross-modal retrieval on RSITMD with 67.44%(I→T) and 74.76%(T→I) at R@10; and improves DIOR-RSVG visual-grounding precision to 64.21% at Pr@0.9 . These gains matter in real deployments by reducing missed targets and speeding analyst search on resource-constrained platforms.  \nKeywords: remote sensing; knowledge distillation; reinforcement learning; classification; retrieval  \n1. Introduction  \nKnowledge distillation (KD) [1–3] has emerged as a powerful technique for transferring knowledge from large, high-capacity teacher models to compact, efficient student models. By leveraging soft target distributions and feature-based guidance, KD enables smaller models to achieve competitive performance while reducing computational costs. This paradigm is particularly valuable for resource-constrained environments, where deploying large-scale models is impractical.  \nDeep learning models have been increasingly applied to complex domains, where highdimensional, multimodal data pose significant challenges. Remote sensing tasks—such as land cover classification, object detection, and image retrieval—require models capable of handling diverse spatial and spectral features [4–6] . Similarly, medical imaging involves high-stakes decision-making, privacy constraints, and substantial variations in data quality across institutions ","cbCaib1IR9uhiaYs","https://ap.wps.com/l/cbCaib1IR9uhiaYs","pdf",1966150,"English","# Abstract\n# Keywords\n# Introduction","[{\"question\":\"What problem does the proposed KDRL framework address?\",\"answer\":\"It addresses limitations of conventional KD under domain shift and task complexity, aiming to transfer robust teacher representations to a compact student without over-regularization or bias amplification.\"},{\"question\":\"How does the framework work in two stages?\",\"answer\":\"Stage one performs supervised fine-tuning using logit and feature-based distillation. Stage two refines the trained student with reinforcement learning using confidence-based and cluster alignment rewards while dynamically reducing reliance on task loss.\"},{\"question\":\"What key innovation is introduced to improve feature learning?\",\"answer\":\"The student encoder includes auxiliary layers that evaluate and reward alignment between learned characteristics and the teacher’s cluster centers, promoting robust feature learning.\"}]","Knowledge Distillation Meets Reinforcement Learning: A Cluster-Driven Approach to Image Processing | PDF",1790767766,48,[146,150,154,158],{"page_no":8,"url":147,"width":148,"height":149},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/2bd7029a77707e3efd608f864f1c0c87.webp",983,1390,{"page_no":14,"url":151,"width":152,"height":153},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/dc8a90ceb2dca89683a885f54a0bedec.webp",982,1388,{"page_no":81,"url":155,"width":156,"height":157},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/71a7f1a0a6c31b3f419a6b1048e4217a.webp",981,1387,{"page_no":19,"url":159,"width":160,"height":161},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/c8e8f8b6680c6672cd12610388134397.webp",980,1385]