{"id":"https://openalex.org/W7171375993","doi":"https://doi.org/10.48550/arxiv.2607.22389","title":"HiKV: Hierarchical Importance-Aware KV Cache with Hardware Acceleration for LLM Decoding","display_name":"HiKV: Hierarchical Importance-Aware KV Cache with Hardware Acceleration for LLM Decoding","publication_year":2026,"publication_date":"2026-07-24","ids":{"openalex":"https://openalex.org/W7171375993","doi":"https://doi.org/10.48550/arxiv.2607.22389"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.22389","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.22389","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2607.22389","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5143730417","display_name":"Chao Fang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Chao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003134609","display_name":"Yin Jz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yin, Jun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043501540","display_name":"Man Shi","orcid":"https://orcid.org/0000-0003-1675-6235"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Man","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5012150553","display_name":"Marian Verhelst","orcid":"https://orcid.org/0000-0003-3495-9263"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Verhelst, Marian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.6276999711990356,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.6276999711990356,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.08860000222921371,"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/T14347","display_name":"Big Data and Digital Economy","score":0.07919999957084656,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.8241999745368958},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.678600013256073},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5972999930381775},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.5428000092506409},{"id":"https://openalex.org/keywords/cpu-cache","display_name":"CPU cache","score":0.4648999869823456},{"id":"https://openalex.org/keywords/sorting","display_name":"Sorting","score":0.46219998598098755},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4474000036716461},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.4284999966621399},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3953999876976013}],"concepts":[{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.8241999745368958},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8014000058174133},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.678600013256073},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6157000064849854},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5972999930381775},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5428000092506409},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.5386000275611877},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.4648999869823456},{"id":"https://openalex.org/C111696304","wikidata":"https://www.wikidata.org/wiki/Q2303697","display_name":"Sorting","level":2,"score":0.46219998598098755},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4474000036716461},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.4284999966621399},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.42719998955726624},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3953999876976013},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.3562000095844269},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.34599998593330383},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.3352000117301941},{"id":"https://openalex.org/C68043766","wikidata":"https://www.wikidata.org/wiki/Q267416","display_name":"Static random-access memory","level":2,"score":0.32179999351501465},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3206000030040741},{"id":"https://openalex.org/C113166858","wikidata":"https://www.wikidata.org/wiki/Q5015981","display_name":"Cache pollution","level":5,"score":0.3197000026702881},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.3093999922275543},{"id":"https://openalex.org/C177950962","wikidata":"https://www.wikidata.org/wiki/Q10997658","display_name":"Non-volatile memory","level":2,"score":0.2856000065803528},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.28360000252723694},{"id":"https://openalex.org/C64540648","wikidata":"https://www.wikidata.org/wiki/Q646477","display_name":"Sorting network","level":4,"score":0.2793000042438507},{"id":"https://openalex.org/C25536678","wikidata":"https://www.wikidata.org/wiki/Q5015977","display_name":"Cache invalidation","level":5,"score":0.2777999937534332},{"id":"https://openalex.org/C167713795","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"Smart Cache","level":5,"score":0.275299996137619},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.274399995803833},{"id":"https://openalex.org/C201148951","wikidata":"https://www.wikidata.org/wiki/Q5015976","display_name":"Cache coloring","level":4,"score":0.2660999894142151},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.2533999979496002},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.25270000100135803}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.22389","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.22389","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2607.22389","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.22389","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy","score":0.8003110885620117}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"With":[0],"the":[1,10,19,46,67,97,105,129,133,174],"rapid":[2],"adoption":[3],"of":[4,70],"long-context":[5],"large":[6],"language":[7],"models":[8],"(LLMs),":[9],"continuously":[11],"growing":[12],"KV":[13,36,47,135],"cache":[14,37,48,136],"during":[15],"decoding":[16],"has":[17],"become":[18],"critical":[20],"memory":[21,159],"bottleneck.":[22],"To":[23],"tackle":[24],"this":[25],"challenge,":[26],"we":[27,83],"propose":[28],"HiKV,":[29],"a":[30,58,79,85,90],"novel":[31],"algorithm-hardware":[32],"co-design":[33],"that":[34,94,169],"exploits":[35],"redundancy":[38],"through":[39],"hierarchical":[40],"importance":[41,92],"awareness.":[42],"Algorithmically,":[43],"HiKV":[44,118,146],"compresses":[45],"at":[49,78],"two":[50],"granularities:":[51],"Stage":[52,62],"I":[53],"evicts":[54],"unimportant":[55],"tokens":[56],"within":[57,138],"fixed":[59],"budget,":[60],"and":[61,124],"II":[63],"further":[64],"loads":[65],"only":[66,171],"significant":[68],"elements":[69],"each":[71,101],"retained":[72],"token,":[73],"reaching":[74],"compression":[75],"ratios":[76],"unattainable":[77],"single":[80],"granularity.":[81],"Architecturally,":[82],"develop":[84],"dedicated":[86],"accelerator":[87],"centered":[88],"on":[89,115],"reconfigurable":[91],"sorter":[93],"switches":[95],"between":[96],"distinct":[98],"sorting":[99],"datapaths":[100],"stage":[102],"requires,":[103],"unifying":[104],"two-stage":[106],"acceleration":[107],"in":[108,128,157],"one":[109],"circuit":[110],"with":[111],"minimal":[112],"overhead.":[113],"Evaluated":[114],"representative":[116],"LLMs,":[117],"achieves":[119],"up":[120],"to":[121,173],"7.95x":[122],"speedup":[123],"90%":[125],"energy":[126],"reduction":[127,156],"attention":[130],"computation":[131],"over":[132],"vanilla":[134],"baseline":[137],"negligible":[139],"1%":[140],"accuracy":[141],"loss.":[142],"Under":[143],"iso-accuracy":[144],"constraints,":[145],"outperforms":[147],"state-of-the-art":[148],"importance-based":[149],"methods":[150],"by":[151,165],"achieving":[152],"an":[153],"additional":[154],"1.82~4.87x":[155],"external":[158],"accesses.":[160],"These":[161],"benefits":[162],"are":[163],"enabled":[164],"specialized":[166],"hardware":[167],"components":[168],"add":[170],"8%":[172],"system":[175],"area.":[176]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-28T00:00:00"}