{"id":"https://openalex.org/W7166880899","doi":"https://doi.org/10.18653/v1/2026.acl-long.326","title":"MixKVQ: Query-Aware Mixed-Precision KV Cache Quantization for Long-Context Reasoning","display_name":"MixKVQ: Query-Aware Mixed-Precision KV Cache Quantization for Long-Context Reasoning","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166880899","doi":"https://doi.org/10.18653/v1/2026.acl-long.326"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.326","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.326","pdf_url":"https://aclanthology.org/2026.acl-long.326.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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.326.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139850041","display_name":"Tao Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001026113","display_name":"Ziqian Zeng","orcid":"https://orcid.org/0000-0003-0734-0035"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ziqian Zeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139741252","display_name":"Hao Peng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hao Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139800289","display_name":"Huiping Zhuang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huiping Zhuang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139825143","display_name":"Cen Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cen Chen","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.87271561,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"7189","last_page":"7204"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11106","display_name":"Data Management and Algorithms","score":0.09200000017881393,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11106","display_name":"Data Management and Algorithms","score":0.09200000017881393,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10444","display_name":"Context-Aware Activity Recognition Systems","score":0.08460000157356262,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11478","display_name":"Caching and Content Delivery","score":0.07150000333786011,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.4081999957561493},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.3199000060558319},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.2775000035762787},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.26829999685287476}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6322000026702881},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4878000020980835},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.4081999957561493},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.3199000060558319},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.3116999864578247},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3003999888896942},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.29089999198913574},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.2775000035762787},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.26829999685287476},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.25360000133514404}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.326","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.326","pdf_url":"https://aclanthology.org/2026.acl-long.326.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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.326","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.326","pdf_url":"https://aclanthology.org/2026.acl-long.326.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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2287814283","display_name":null,"funder_award_id":"62472181","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G251544062","display_name":null,"funder_award_id":"2024ZYGXZR074","funder_id":"https://openalex.org/F4320335787","funder_display_name":"Fundamental Research Funds for the Central Universities"},{"id":"https://openalex.org/G4421831070","display_name":null,"funder_award_id":"62406114","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G645884509","display_name":null,"funder_award_id":"2024A1515010220","funder_id":"https://openalex.org/F4320337111","funder_display_name":"Basic and Applied Basic Research Foundation of Guangdong Province"},{"id":"https://openalex.org/G7626417556","display_name":null,"funder_award_id":"62306117","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G77119337","display_name":null,"funder_award_id":"2024D03J0005","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320318547","display_name":"Baidu","ror":"https://ror.org/03vs3wt56"},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335787","display_name":"Fundamental Research Funds for the Central Universities","ror":null},{"id":"https://openalex.org/F4320337111","display_name":"Basic and Applied Basic Research Foundation of Guangdong Province","ror":null}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166880899.pdf","grobid_xml":"https://content.openalex.org/works/W7166880899.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Long":[0],"Chain-of-Thought":[1],"(CoT)":[2],"reasoning":[3,50,135],"has":[4],"significantly":[5,141],"advanced":[6],"the":[7,26,59,97],"capabilities":[8],"of":[9],"Large":[10],"Language":[11],"Models":[12],"(LLMs),":[13],"but":[14],"this":[15,100],"progress":[16],"is":[17,34],"accompanied":[18],"by":[19],"substantial":[20],"memory":[21,157],"and":[22,93,117],"latency":[23],"overhead":[24],"from":[25],"extensive":[27],"Key-Value":[28],"(KV)":[29],"cache.Although":[30],"KV":[31,79],"cache":[32,80],"quantization":[33,41,52,81,91,107,129],"a":[35,87,105,111,150,154],"promising":[36],"compression":[37],"technique,":[38],"existing":[39,143],"low-bit":[40,78,144],"methods":[42],"often":[43],"exhibit":[44],"severe":[45],"performance":[46,147],"degradation":[47],"on":[48,99,133],"complex":[49,134],"tasks.Fixed-precision":[51],"struggles":[53],"to":[54,67,96,115,149],"handle":[55],"outlier":[56],"channels":[57,121],"in":[58],"key":[60,88,120],"cache,":[61],"while":[62,126],"current":[63],"mixed-precision":[64],"strategies":[65],"fail":[66],"accurately":[68],"identify":[69,116],"components":[70],"requiring":[71],"high-precision":[72],"representation.We":[73],"find":[74],"that":[75,109,122,138],"an":[76],"effective":[77],"strategy":[82],"must":[83],"consider":[84],"two":[85],"factors:":[86],"channel's":[89],"intrinsic":[90],"difficulty":[92],"its":[94],"relevance":[95],"query.Based":[98],"insight,":[101],"we":[102],"propose":[103],"MixKVQ,":[104],"novel":[106],"method":[108],"introduces":[110],"lightweight,":[112],"query-aware":[113],"algorithm":[114],"preserve":[118],"critical":[119],"need":[123],"higher":[124],"precision,":[125],"applying":[127],"per-token":[128],"for":[130],"value":[131],"cache.Experiments":[132],"datasets":[136],"demonstrate":[137],"our":[139],"approach":[140],"outperforms":[142],"methods,":[145],"achieving":[146],"comparable":[148],"fullprecision":[151],"baseline":[152],"at":[153],"substantially":[155],"reduced":[156],"footprint.":[158]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}