[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84461-en":3,"doc-seo-84461-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84461,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1782698725881665579",8,"Research & Report","FedMosaic Federated Retrieval-Augmented Generation via Parametric Adapters","Retrieval-Augmented Generation (RAG) improves Large Language Models by grounding responses in external knowledge to enhance factuality and reduce hallucinations, yet centralized corpora are impractical in privacy- and compliance-sensitive domains. Federated RAG (FedRAG) enables collaboration between a central LLM server and distributed silos without sharing raw documents. This work introduces FedMosaic, a federated framework using parametric adapters, addressing adapter storage/communication overhead and destructive aggregation. It clusters semantically related documents into multi-document adapters with document-specific masks and selectively aggregates only relevance-aligned, nonconflicting adapters, achieving higher accuracy while reducing costs.","FedMosaic: Federated Retrieval-Augmented Generation via  \nParametric Adapters  \nZhilin Liang  \nSKLCCSE Lab Beihang University Beijing, China [zlliang@buaa.edu.cn](zlliang@buaa.edu.cn)  \nHainan Zhang  \nBeijing Advanced Innovation Center Beihang University Beijing, China [zhanghainan1990@163.com](zhanghainan1990@163.com)  \nYuxiang Wang  \nSKLCCSE Lab Beihang University Beijing, China [yuxiangwang@buaa.edu.cn](yuxiangwang@buaa.edu.cn)  \nBoyi Liu  \nSKLCCSE Lab Beihang University Beijing, China [boyliu@buaa.edu.cn](boyliu@buaa.edu.cn)  \nZimu Zhou  \nDepartment of Data Science City University of Hong Kong Hong Kong, China[zimuzhou@cityu.edu.hk](zimuzhou@cityu.edu.hk)  \nYongxin Tong∗ SKLCCSE Lab Beihang University Beijing, China [yxtong@buaa.edu.cn](yxtong@buaa.edu.cn)  \narXiv :2602 .05235v2 [ cs .CL] 11 Jul 2026  \nAbstract  \nRetrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by grounding generation in external knowledge to improve factuality and reduce hallucinations. Yet most deployments assume a centralized corpus, which is infeasible in privacyaware domains where knowledge remains siloed. This motivates federated RAG (FedRAG), where a central LLM server collaborates with distributed silos without sharing raw documents. Incontext RAG violates this requirement by transmitting verbatim documents, whereas parametric RAG encodes documents into lightweight adapters that merge with a frozen LLM at inference, avoiding raw-text exchange. We adopt the parametric approach but face two unique challenges induced by FedRAG: high storage and communication from per-document adapters, and destructive aggregation caused by indiscriminately merging multiple adapters. We present FedMosaic, the first federated RAG framework built on parametric adapters. FedMosaic clusters semantically related documents into multi-document adapters with document-specific masks to reduce overhead while preserving specificity, and performs selective adapter aggregation to combine only relevance-aligned, nonconflicting adapters. Experiments show that FedMosaic achievesan average 10. 9% higher accuracy than state-of-the-art methods in four categories, while lowering storage costs by 78. 8% to 86. 3% and communication costs by 91.4%, and never sharing raw documents1 .  \nCCS Concepts  \n• Computing methodologies → Learning paradigms.  \nKeywords  \nRetrieval-Augmented Generation; Parametric Adapters; Federated Computing  \n∗ Corresponding author.  \n1 We have open-sourced the code at: [https://github.com/lewellin727/FedMosaic](https://github.com/lewellin727/FedMosaic)  \nThis work is licensed under a Creative Commons Attribution 4 .0 International License. SIGIR ’26, Melbourne, VIC, Australia  \n© 2026 Copyright held by the owner/author(s) .  \nACM ISBN 979-8-4007-2599-9/2026/07  \n[https://doi.org/10.1145/3805712.3809655](https://doi.org/10.1145/3805712.3809655)  \nACM Reference Format:  \nZhilin Liang, Yuxiang Wang, Zimu Zhou, Hainan Zhang, Boyi Liu, and Yongxin Tong. 2026. FedMosaic: Federated Retrieval-Augmented Generation via Parametric Adapters. In Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR’26), July 20–24, 2026, Melbourne, VIC, Australia. ACM, New York, NY, USA, 11 pages. [https://doi.org/10.1145/3805712.3809655](https://doi.org/10.1145/3805712.3809655)  \n1 Introduction  \nRetrieval-Augmented Generation (RAG) [12] has emerged a core technique for enhancing Large Language Models (LLMs) by grounding their responses in dynamic external knowledge rather than static pre-training alone. By retrieving relevant documents and conditioning generation on them, RAG improves factual accuracy and mitigates hallucination, driving its adoption in applications such as conversational search [43], enterprise knowledge assistants [6], and domain-specific Q&A [50] . However, most RAG deployments assume access to a centralized corpus (e.g. Wikipedia, Common Crawl, or enterprise repositories). While feasi","cbCaiqflBJziKTUF","https://ap.wps.com/l/cbCaiqflBJziKTUF","pdf",1137751,4,1,11,"English","en",105,"# Introduction\n## Problem: RAG locality constraint in federated domains\n## Motivation: privacy-aware collaboration across silos\n## Limitations of conventional and prompt-noising approaches","[{\"question\":\"What problem does FedMosaic address in federated RAG?\",\"answer\":\"FedMosaic targets federated RAG scenarios where raw documents cannot be shared due to the locality constraint, making conventional prompt-based RAG infeasible. It also addresses high storage/communication overhead from per-document adapters and destructive effects from indiscriminate adapter merging.\"},{\"question\":\"How does FedMosaic use parametric adapters to avoid sharing raw documents?\",\"answer\":\"FedMosaic encodes documents into lightweight parametric adapters that are merged with a frozen LLM during inference. This avoids transmitting verbatim documents from silos to the central server while still leveraging external knowledge.\"},{\"question\":\"How does FedMosaic reduce adapter overhead and prevent harmful aggregation?\",\"answer\":\"FedMosaic clusters semantically related documents into multi-document adapters using document-specific masks to reduce overhead while keeping specificity. It then performs selective adapter aggregation, combining only relevance-aligned and nonconflicting adapters to avoid destructive aggregation.\"}]",1784195780,28,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"fedmosaic-federated-retrieval-augmented-generation-via-parametric-adapters","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/fedmosaic-federated-retrieval-augmented-generation-via-parametric-adapters/84461/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-21","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does FedMosaic address in federated RAG?","Question",{"text":75,"@type":76},"FedMosaic targets federated RAG scenarios where raw documents cannot be shared due to the locality constraint, making conventional prompt-based RAG infeasible. It also addresses high storage/communication overhead from per-document adapters and destructive effects from indiscriminate adapter merging.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does FedMosaic use parametric adapters to avoid sharing raw documents?",{"text":80,"@type":76},"FedMosaic encodes documents into lightweight parametric adapters that are merged with a frozen LLM during inference. This avoids transmitting verbatim documents from silos to the central server while still leveraging external knowledge.",{"name":82,"@type":73,"acceptedAnswer":83},"How does FedMosaic reduce adapter overhead and prevent harmful aggregation?",{"text":84,"@type":76},"FedMosaic clusters semantically related documents into multi-document adapters using document-specific masks to reduce overhead while keeping specificity. It then performs selective adapter aggregation, combining only relevance-aligned and nonconflicting adapters to avoid destructive aggregation.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]