{"id":"https://openalex.org/W7133489089","doi":"https://doi.org/10.1109/tcsi.2026.3668423","title":"CrossKV: Accelerating Large Language Model Inference via Cross-Stage Dynamic Co-Optimization for KV Cache","display_name":"CrossKV: Accelerating Large Language Model Inference via Cross-Stage Dynamic Co-Optimization for KV Cache","publication_year":2026,"publication_date":"2026-03-04","ids":{"openalex":"https://openalex.org/W7133489089","doi":"https://doi.org/10.1109/tcsi.2026.3668423"},"language":null,"primary_location":{"id":"doi:10.1109/tcsi.2026.3668423","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2026.3668423","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems I: Regular Papers","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128121803","display_name":"Nan Wu","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Nan Wu","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0005-2703-9938","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052871033","display_name":"Siqi Wang","orcid":"https://orcid.org/0009-0006-9675-7973"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shenyu Wang","raw_affiliation_strings":["College of Biomedical Engineering, Fudan University, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0008-2549-6264","affiliations":[{"raw_affiliation_string":"College of Biomedical Engineering, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089405414","display_name":"Huizheng Wang","orcid":"https://orcid.org/0000-0002-9763-8208"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huizheng Wang","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-9763-8208","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5146914236","display_name":"Peng Wang","orcid":"https://orcid.org/0000-0002-2631-971X"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peng Wang","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-2631-971X","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100441023","display_name":"Xiao Liu","orcid":"https://orcid.org/0000-0001-5514-021X"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiao Liu","raw_affiliation_strings":["College of Biomedical Engineering, Fudan University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0001-5514-021X","affiliations":[{"raw_affiliation_string":"College of Biomedical Engineering, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128094738","display_name":"Zhihua Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhihua Wang","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-6567-0759","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066434078","display_name":"Yang Hu","orcid":"https://orcid.org/0000-0001-6942-4395"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang Hu","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-6942-4395","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5063888012","display_name":"Hanjun Jiang","orcid":"https://orcid.org/0000-0003-4911-0748"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hanjun Jiang","raw_affiliation_strings":["School of Integrated Circuits, Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-4911-0748","affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":{"value":2645,"currency":"USD","value_usd":2645},"apc_paid":null,"fwci":7.0482,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.95179317,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"73","issue":"7","first_page":"4801","last_page":"4814"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.3041999936103821,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.3041999936103821,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.2354000061750412,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.053700000047683716,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.5824999809265137},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4832000136375427},{"id":"https://openalex.org/keywords/data-modeling","display_name":"Data modeling","score":0.36149999499320984},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.33719998598098755}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7718999981880188},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5824999809265137},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4832000136375427},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.39989998936653137},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.3630000054836273},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.36149999499320984},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.33719998598098755},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3215999901294708},{"id":"https://openalex.org/C24326235","wikidata":"https://www.wikidata.org/wiki/Q126095","display_name":"Electronic engineering","level":1,"score":0.29179999232292175},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.26649999618530273},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2644999921321869}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tcsi.2026.3668423","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2026.3668423","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems I: Regular Papers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G427040061","display_name":null,"funder_award_id":"XMHT20220115007","funder_id":"https://openalex.org/F4320327511","funder_display_name":"Development and Reform Commission of Shenzhen Municipality"},{"id":"https://openalex.org/G4401221671","display_name":null,"funder_award_id":"2021Szvup094","funder_id":"https://openalex.org/F4320329836","funder_display_name":"Chuzhou Science and Technology Program"},{"id":"https://openalex.org/G4700936218","display_name":null,"funder_award_id":"GJHZ20210705142401005","funder_id":"https://openalex.org/F4320329836","funder_display_name":"Chuzhou Science and Technology Program"},{"id":"https://openalex.org/G5003656628","display_name":null,"funder_award_id":"2023YFC3604201","funder_id":"https://openalex.org/F4320327808","funder_display_name":"Yunnan Key Research and Development Program"},{"id":"https://openalex.org/G5069898323","display_name":null,"funder_award_id":"JCYJ20190807142805490","funder_id":"https://openalex.org/F4320329836","funder_display_name":"Chuzhou Science and Technology Program"},{"id":"https://openalex.org/G5469626880","display_name":null,"funder_award_id":"2023YFC3604200","funder_id":"https://openalex.org/F4320327808","funder_display_name":"Yunnan Key Research and Development Program"}],"funders":[{"id":"https://openalex.org/F4320327511","display_name":"Development and Reform Commission of Shenzhen Municipality","ror":"https://ror.org/03jmg4515"},{"id":"https://openalex.org/F4320327808","display_name":"Yunnan Key Research and Development Program","ror":null},{"id":"https://openalex.org/F4320329836","display_name":"Chuzhou Science and Technology Program","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":23,"referenced_works":["https://openalex.org/W2040903332","https://openalex.org/W2798405286","https://openalex.org/W2923014074","https://openalex.org/W2979826702","https://openalex.org/W3159727696","https://openalex.org/W3190062760","https://openalex.org/W3205192296","https://openalex.org/W4214686755","https://openalex.org/W4313467238","https://openalex.org/W4360831786","https://openalex.org/W4360831803","https://openalex.org/W4360832001","https://openalex.org/W4393152626","https://openalex.org/W4393407021","https://openalex.org/W4396909834","https://openalex.org/W4399146344","https://openalex.org/W4401211590","https://openalex.org/W4401247357","https://openalex.org/W4404954358","https://openalex.org/W4406521183","https://openalex.org/W4409248488","https://openalex.org/W4413966866","https://openalex.org/W7092191477"],"related_works":[],"abstract_inverted_index":{"The":[0],"autoregressive":[1],"nature":[2],"of":[3,55,71,100,169,173],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"has":[8],"enabled":[9],"remarkable":[10],"performance":[11],"in":[12,19,68,113],"language":[13,21],"generation,":[14],"making":[15],"them":[16],"a":[17,43,69,87,118,123,129,135],"cornerstone":[18],"natural":[20],"processing.":[22],"As":[23],"context":[24],"windows":[25],"lengthen,":[26],"the":[27,39,98,110,143,167,203],"per-token":[28],"key\u2013value":[29],"(KV)":[30],"cache":[31,57],"grows":[32],"linearly":[33],"with":[34,158],"sequence":[35],"length":[36],"and":[37,59,74,116,134,180,193],"turns":[38],"generation":[40],"stage":[41,54],"into":[42],"memory-bound":[44],"operation.":[45],"However,":[46],"existing":[47,197],"works":[48],"focus":[49],"primarily":[50],"on":[51],"an":[52,151,185,209],"isolated":[53],"KV":[56,101,114,144],"reduction,":[58],"most":[60],"efforts":[61],"incur":[62],"significant":[63,76],"hardware":[64,160],"overheads.":[65],"This":[66],"results":[67],"lack":[70],"cross-stage":[72,119,155],"co-optimization":[73,120,157],"leaves":[75],"reduction":[77],"potential":[78,99],"untapped.":[79],"To":[80],"address":[81],"these":[82],"challenges,":[83],"we":[84,108],"propose":[85],"CrossKV,":[86],"software-hardware":[88],"co-design":[89],"architecture":[90,153],"that":[91],"accelerates":[92],"LLM":[93],"inference":[94],"to":[95,141,164,196,202],"fully":[96,165],"achieve":[97],"caching":[102,115,145],"optimization.":[103],"Motivated":[104],"by":[105],"our":[106],"observation,":[107],"identify":[109],"three":[111],"stages":[112],"present":[117],"algorithm,":[121],"including:":[122],"derivative-enhanced":[124],"dynamic":[125,136,156],"progressive":[126],"pruning":[127],"method,":[128],"DCT-driven":[130],"key-vector":[131],"low-rank":[132],"compressing,":[133],"clustered":[137],"hybrid":[138],"run-length":[139],"encoding":[140],"reduce":[142],"while":[146],"maintaining":[147],"high":[148],"accuracy.":[149],"Then,":[150],"efficient":[152],"featuring":[154],"negligible":[159],"overhead":[161],"is":[162],"proposed":[163],"harness":[166],"algorithm":[168],"CrossKV.":[170],"Our":[171],"evaluations":[172],"CrossKV":[174,207],"over":[175],"9":[176],"popular":[177],"LLMs":[178],"models":[179],"various":[181],"long-context":[182],"tasks":[183],"demonstrate":[184],"average":[186],"of$6.03\\times":[187],"$and$4.10\\times":[188],"$improvement":[189],"for":[190],"energy":[191],"efficiency":[192,212],"speedup":[194],"compared":[195],"SoTA":[198],"Accelerators":[199],"architecture.":[200],"Compared":[201],"Nvidia":[204],"A100":[205],"GPU,":[206],"achieves":[208],"average$23.43\\times":[210],"$energy":[211],"and$4.64\\times":[213],"$speedup,":[214],"respectively.":[215]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-08-21T09:56:20.448147","created_date":"2026-03-05T00:00:00"}
