{"id":"https://openalex.org/W4416035984","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1079","title":"TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection","display_name":"TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416035984","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1079"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1079","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1079","pdf_url":"https://aclanthology.org/2025.emnlp-main.1079.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.1079.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101632556","display_name":"Wei Wu","orcid":"https://orcid.org/0000-0002-9624-2949"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Wei Wu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014869504","display_name":"Zhuoshi Pan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhuoshi Pan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011657849","display_name":"Kun Fu","orcid":"https://orcid.org/0000-0003-4963-615X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kun Fu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100406893","display_name":"Chao Wang","orcid":"https://orcid.org/0000-0001-7717-447X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chao Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056428506","display_name":"Liyi Chen","orcid":"https://orcid.org/0000-0003-2166-4386"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liyi Chen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087340354","display_name":"Y BAI","orcid":"https://orcid.org/0009-0006-2459-5530"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yunchu Bai","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102738457","display_name":"Tianfu Wang","orcid":"https://orcid.org/0000-0002-4386-3354"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianfu Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082429626","display_name":"Zheng Wang","orcid":"https://orcid.org/0000-0001-8305-0676"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5085052668","display_name":"Hui Xiong","orcid":"https://orcid.org/0000-0003-4138-2992"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hui Xiong","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5101632556"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.7531,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":{"value":0.92471399,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":91,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"21275","last_page":"21292"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.4065000116825104,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.4065000116825104,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.14560000598430634,"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/T12326","display_name":"Network Packet Processing and Optimization","score":0.13169999420642853,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.6503000259399414},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.5976999998092651},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.573199987411499},{"id":"https://openalex.org/keywords/extrapolation","display_name":"Extrapolation","score":0.5536999702453613}],"concepts":[{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.6503000259399414},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6381000280380249},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5976999998092651},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.573199987411499},{"id":"https://openalex.org/C132459708","wikidata":"https://www.wikidata.org/wiki/Q744069","display_name":"Extrapolation","level":2,"score":0.5536999702453613},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4032999873161316},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3806999921798706},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2678999900817871},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.263700008392334},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.262800008058548},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2614000141620636}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.1079","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1079","pdf_url":"https://aclanthology.org/2025.emnlp-main.1079.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1079","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1079","pdf_url":"https://aclanthology.org/2025.emnlp-main.1079.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1121271761","display_name":null,"funder_award_id":"Program","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G1477544716","display_name":null,"funder_award_id":"Guangdong","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G1782397203","display_name":null,"funder_award_id":"92370204","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2087396116","display_name":null,"funder_award_id":"China","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3085993365","display_name":null,"funder_award_id":"(Grant No.","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3317480652","display_name":null,"funder_award_id":"Science","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G37568934","display_name":null,"funder_award_id":"Grant","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5957623925","display_name":null,"funder_award_id":"2023B151","funder_id":"https://openalex.org/F4320337111","funder_display_name":"Basic and Applied Basic Research Foundation of Guangdong Province"},{"id":"https://openalex.org/G5994120800","display_name":null,"funder_award_id":"Natural","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G6130982387","display_name":null,"funder_award_id":"2023B1515120057","funder_id":"https://openalex.org/F4320337111","funder_display_name":"Basic and Applied Basic Research Foundation of Guangdong Province"},{"id":"https://openalex.org/G7726157001","display_name":null,"funder_award_id":"Grant No.","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320337111","display_name":"Basic and Applied Basic Research Foundation of Guangdong Province","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416035984.pdf","grobid_xml":"https://content.openalex.org/works/W4416035984.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Rapid":[0],"advances":[1],"in":[2,15,46,101,144,152],"Large":[3],"Language":[4],"Models":[5],"(LLMs)":[6],"have":[7],"spurred":[8],"demand":[9],"for":[10,62],"processing":[11],"extended":[12],"context":[13],"sequences":[14],"contemporary":[16],"applications.However,":[17],"this":[18,49],"progress":[19],"faces":[20],"two":[21],"challenges:":[22],"performance":[23,158],"degradation":[24],"due":[25],"to":[26,80,141,149,160],"sequence":[27],"lengths":[28],"out-of-distribution,":[29],"and":[30,64,122,147],"excessively":[31],"long":[32],"inference":[33,163],"times":[34],"caused":[35],"by":[36],"the":[37,70,112,124,132],"quadratic":[38],"computational":[39],"complexity":[40],"of":[41,72,118,137],"attention.These":[42],"issues":[43],"limit":[44],"LLMs":[45],"long-context":[47,66,162],"scenarios.In":[48],"paper,":[50],"we":[51,110],"propose":[52],"Dynamic":[53],"Token-Level":[54],"KV":[55,83,98],"Cache":[56,84,114],"Selection":[57,113],"(TokenSelect),":[58],"a":[59,95],"training-free":[60],"method":[61],"efficient":[63,125],"accurate":[65],"inference.TokenSelect":[67],"builds":[68],"upon":[69],"observation":[71],"non-contiguous":[73],"attention":[74,102,145],"sparsity,":[75],"using":[76],"QK":[77],"dot":[78],"products":[79],"measure":[81],"per-head":[82,88],"criticality":[85],"at":[86],"tokenlevel.By":[87],"soft":[89],"voting":[90],"mechanism,":[91],"To-kenSelect":[92,138],"selectively":[93],"involves":[94],"few":[96],"critical":[97],"cache":[99],"tokens":[100],"calculation":[103],"without":[104],"sacrificing":[105],"accuracy.To":[106],"further":[107],"accelerate":[108],"To-kenSelect,":[109],"design":[111],"based":[115],"on":[116],"observations":[117],"consecutive":[119],"Query":[120],"similarity":[121],"implemented":[123],"Paged":[126],"Dot":[127],"Product":[128],"Kernel,":[129],"significantly":[130],"reducing":[131],"selection":[133],"overhead.A":[134],"comprehensive":[135],"evaluation":[136],"demonstrates":[139],"up":[140,148],"23.84":[142],"speedup":[143],"computation":[146],"2.28":[150],"acceleration":[151],"end-to-end":[153],"latency,":[154],"while":[155],"providing":[156],"superior":[157],"compared":[159],"state-of-theart":[161],"methods.":[164]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-04-18T07:56:08.524223","created_date":"2025-11-08T00:00:00"}