[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273660-105":59,"doc-detail-273660-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","recache-efficient-kv-cache-reuse-and-compression-for-tool-augmented-llm-agents","ReCache - Efficient KV Cache Reuse and Compression for Tool-Augmented LLM Agents","","Agentic language models repeatedly encode tool and skill schemas that recur across requests in different combinations and orders, preventing standard prefix caching from reusing KV states. ReCache introduces independent caching for reusable resource representations, reducing inference-time computation and memory via resource-wise attention, which removes cross-resource interactions and yields composition-invariant KV blocks. It further prunes structural and semantic visibility along contribution-selected routes. Experiments on a benchmark from seven public datasets show dense performance parity and a 3.655× time-to-first-token speedup with large KV memory reductions.",{"@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":30,"@type":76,"position":81},"https://docshare.wps.com/document/technology/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/recache-efficient-kv-cache-reuse-and-compression-for-tool-augmented-llm-agents/273660/",{"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/recache-efficient-kv-cache-reuse-and-compression-for-tool-augmented-llm-agents/273660.png","ImageObject",300,407,{"name":92,"@type":93},"Riley","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-09","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":52},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why does standard prefix caching fail for tool-augmented LLM agents?","Question",{"text":112,"@type":113},"Agentic requests dynamically compose tool and skill schemas in different combinations and orders, so reused content rarely forms an identical prefix. This prevents reuse of previously computed KV states.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What is resource-wise attention in ReCache?",{"text":117,"@type":113},"ReCache uses resource-wise attention to remove attention between distinct resource blocks and reset positional indices within each block. The resulting KV representation becomes independent of neighboring resources and their absolute positions.",{"name":119,"@type":110,"acceptedAnswer":120},"How does ReCache reduce memory and computation beyond independent KV reuse?",{"text":121,"@type":113},"ReCache compresses resource caches structurally by retaining only contribution-selected layer–KV-head-group routes, and semantically by pruning fields while keeping invocation-critical information.","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},273660,1790307493,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":29,"category_name":30,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":52,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145},1374391975076,"https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051","ReCache: Efficient KV Cache Reuse and Compression for Tool-Augmented  \nLLM Agents  \nYichu Fang1,2 , Sitong Wei3 , Haozhe Hu1,2 , Xiaoyu Shen2 *  \n1 Shanghai Jiao Tong University  \n2Eastern Institute of Technology, Ningbo  \n3Xi’an Jiaotong University  \n[fy_chu@sjtu.edu.cn xyshen@eitech.edu.cn](fy_chu@sjtu.edu.cn xyshen@eitech.edu.cn)  \narXiv : 2608 . 19662v 1 [cs .CL] 20 Aug 2026  \nAbstract  \nAgentic language models repeatedly encode tool and skill schemas that recur across requests in different combinations and orders, preventing standard prefix caching from reusing their key–value (KV) states. We introduce ReCache, a framework for independently caching resource representations while reducing their inference-time computational and memory overhead. Resource-wise attention removes cross-resource interactions and assigns resource-local positions, producing composition-invariant KV blocks. ReCache then restricts resource visibility to contributionselected layer–KV-head-group routes and retains only invocation-critical fields through structural and semantic pruning. We evaluate ReCache on a benchmark assembled from seven public tool-and skill-use datasets, including resource-disjoint tests. Resource-wise attention matches dense invocation performance (82.3% versus 82.4% Inv-F1) while providing a 3.655 × time-to-first-token speedup. The complete framework reduces allocated KVtensor memory by 92.43% and accelerates attention by 1.423 × . These results show that separating reusable schema encoding from selective resource access substantially reduces agentic inference costs with limited effectiveness loss. The code is available at [https:](https:)//[github.com/EIT-NLP/ReCache](github.com/EIT-NLP/ReCache).  \n1 Introduction  \nRecent advances have extended large language models (LLMs) into the core reasoning module of agentic systems, where they address user requests by dynamically invoking external tools and skills (Schick et al., 2023 ; Liu et al., 2024a ; Li et al., 2026 ; Xu and Yan, 2026) . As the set expands, agentic systems commonly adopt progressive disclosure and retrieve only task-relevant schemas during execution (Shi et al., 2025 ; Zheng et al., 2026) . We  \n* Corresponding author  \nQuery resource / relative token position  \nR1  \nR2  \nR3  \nR4  \nR1 R2 R3 R4  \nKey resource / relative token position  \n0.20  \n0.15  \n0.10  \n0.05  \nMean attent ion mass  \n0.00  \n(a) Causal attention among resources (relative-position bins) . Orange boxes mark cross-resource blocks.  \n\\#1  \nSubsequent conversations  \n0 0.5 1 Relative token position (avg. over resources)  \n(b) Attention from subsequent conversation tokens onto resources, normalized and aligned by relative position.  \nFigure 1: Cross-resource attention on Qwen3-4B.  \nrefer to each tool or skill schema as a resource. Although retrieval limits the active context, the same resource may recur across requests in different combinations and orders, causing repeated prefill computation for unchanged information.  \nKV-cache reuse provides a natural way to amortize this repeated computation. However, standard prefix caching requires reused content to form an identical prefix, which rarely holds for dynamically composed resource contexts (Ye et al., 2025 ; Zhang et al., 2026) . Recent approaches extend KV reuse to modular or position-independent contexts (Gimet al., 2024 ; Hu et al., 2024 ; Yao et al., 2025 ; Yang et al., 2025c), but often require positional correction or selective recomputation to approximate full-  \ncontext KV states. Meanwhile, transformers can jointly encode semantic and positional dependencies (Vaswani et al., 2017 ; Shaw et al., 2018 ; Suet al., 2024), and the importance of positional information varies across tasks, with some studies showing that certain tasks can maintain performance under reduced or altered positional signals (Wanget al., 2021 ; Haviv et al., 2022 ; Kazemnejad et al., 2023) . In resource invocation, preserving resourceinternal semantics may therefore matt","cbCaiecRouxTSELA","https://ap.wps.com/l/cbCaiecRouxTSELA","pdf",652503,17,"English","# Introduction\n## ReCache motivation: KV reuse vs dynamic resource composition\n## Resource-wise attention for composition-invariant KV blocks\n## Contribution-based structural pruning and semantic compression\n## Experimental evaluation on multi-dataset benchmarks","[{\"question\":\"Why does standard prefix caching fail for tool-augmented LLM agents?\",\"answer\":\"Agentic requests dynamically compose tool and skill schemas in different combinations and orders, so reused content rarely forms an identical prefix. This prevents reuse of previously computed KV states.\"},{\"question\":\"What is resource-wise attention in ReCache?\",\"answer\":\"ReCache uses resource-wise attention to remove attention between distinct resource blocks and reset positional indices within each block. The resulting KV representation becomes independent of neighboring resources and their absolute positions.\"},{\"question\":\"How does ReCache reduce memory and computation beyond independent KV reuse?\",\"answer\":\"ReCache compresses resource caches structurally by retaining only contribution-selected layer–KV-head-group routes, and semantically by pruning fields while keeping invocation-critical information.\"}]","ReCache - Efficient KV Cache Reuse and Compression for Tool-Augmented LLM Agents | PDF",1789472411,43]