{"id":"https://openalex.org/W4412888313","doi":"https://doi.org/10.18653/v1/2025.findings-acl.644","title":"A2ATS: Retrieval-Based KV Cache Reduction via Windowed Rotary Position Embedding and Query-Aware Vector Quantization","display_name":"A2ATS: Retrieval-Based KV Cache Reduction via Windowed Rotary Position Embedding and Query-Aware Vector Quantization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888313","doi":"https://doi.org/10.18653/v1/2025.findings-acl.644"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.644","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.644","pdf_url":"https://aclanthology.org/2025.findings-acl.644.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.644.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5086412366","display_name":"Junhui He","orcid":"https://orcid.org/0009-0002-0125-2018"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junhui He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Junna Xing","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junna Xing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100332757","display_name":"Nan Wang","orcid":"https://orcid.org/0000-0002-4780-2022"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nan Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101947314","display_name":"Rui Xu","orcid":"https://orcid.org/0000-0001-9982-5421"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rui Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5061158039","display_name":"Shang-Yu Wu","orcid":"https://orcid.org/0000-0002-1961-143X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shangyu Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100438332","display_name":"Peng Zhou","orcid":"https://orcid.org/0000-0002-2968-6937"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100409453","display_name":"Qiang Liu","orcid":"https://orcid.org/0000-0002-1028-0582"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiang Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101441768","display_name":"Chun Jason Xue","orcid":"https://orcid.org/0000-0002-6431-9868"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chun Jason Xue","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5103211926","display_name":"Qingan Li","orcid":"https://orcid.org/0000-0003-0110-5405"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qingan Li","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":0.1434145,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12451","last_page":"12463"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11269","display_name":"Algorithms and Data Compression","score":0.9955000281333923,"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/T11269","display_name":"Algorithms and Data Compression","score":0.9955000281333923,"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/T12326","display_name":"Network Packet Processing and Optimization","score":0.9894999861717224,"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.9879999756813049,"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/computer-science","display_name":"Computer science","score":0.670067310333252},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.6573723554611206},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.5610055327415466},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.5295584797859192},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5168636441230774},{"id":"https://openalex.org/keywords/position","display_name":"Position (finance)","score":0.4379136562347412},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3971695899963379},{"id":"https://openalex.org/keywords/computer-vision","display_name":"Computer vision","score":0.32328957319259644},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.13570931553840637}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.670067310333252},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.6573723554611206},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.5610055327415466},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.5295584797859192},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5168636441230774},{"id":"https://openalex.org/C198082294","wikidata":"https://www.wikidata.org/wiki/Q3399648","display_name":"Position (finance)","level":2,"score":0.4379136562347412},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3971695899963379},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.32328957319259644},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.13570931553840637},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.0},{"id":"https://openalex.org/C10138342","wikidata":"https://www.wikidata.org/wiki/Q43015","display_name":"Finance","level":1,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.644","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.644","pdf_url":"https://aclanthology.org/2025.findings-acl.644.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},{"id":"pmh:oai:pure.atira.dk:publications/284d962c-d629-461d-a50d-590a56de4900","is_oa":true,"landing_page_url":"https://hdl.handle.net/2031/284d962c-d629-461d-a50d-590a56de4900","pdf_url":null,"source":{"id":"https://openalex.org/S7407055387","display_name":"CityU Scholars","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"info:eu-repo/semantics/conferenceObject"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.644","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.644","pdf_url":"https://aclanthology.org/2025.findings-acl.644.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888313.pdf","grobid_xml":"https://content.openalex.org/works/W4412888313.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2081900870","https://openalex.org/W1539956819","https://openalex.org/W2037549926","https://openalex.org/W2093152993","https://openalex.org/W2345479200","https://openalex.org/W2183306018","https://openalex.org/W1489554298","https://openalex.org/W2378212145","https://openalex.org/W2400327593","https://openalex.org/W3209251257"],"abstract_inverted_index":{"Long":[0],"context":[1,153,183],"large":[2,15],"language":[3],"models":[4],"(LLMs)":[5],"pose":[6],"significant":[7],"challenges":[8],"for":[9,147],"efficient":[10,100],"serving":[11,154,184],"due":[12],"to":[13,40,81,95,177,188],"the":[14,36,91,105,116,134,143],"memory":[16],"footprint":[17],"and":[18,42,58,101,121],"high":[19],"access":[20],"overhead":[21,175],"of":[22,86,104,136],"KV":[23,25,38,74,148],"cache.Retrieval-based":[24],"cache":[26,39,75,149],"reduction":[27,76],"methods":[28,51],"can":[29,165],"mitigate":[30],"these":[31,50,63],"challenges,":[32],"typically":[33],"by":[34,89,186],"offloading":[35],"complete":[37],"CPU":[41],"retrieving":[43],"necessary":[44],"tokens":[45],"on":[46],"demand":[47],"during":[48],"inference.However,":[49],"still":[52],"suffer":[53],"from":[54,119],"unsatisfactory":[55],"accuracy":[56],"degradation":[57,170],"extra":[59],"retrieval":[60,103],"overhead.To":[61],"address":[62],"limitations,":[64],"this":[65],"paper":[66],"proposes":[67],"A":[68,162],"2":[69,78,163],"ATS,":[70],"a":[71,167],"novel":[72],"retrieval-based":[73],"method.A":[77],"ATS":[79,164],"aims":[80],"obtain":[82],"an":[83],"accurate":[84],"approximation":[85,139],"attention":[87,137],"scores":[88],"applying":[90],"vector":[92,130],"quantization":[93,131],"technique":[94],"key":[96,122],"states,":[97],"thereby":[98,180],"enabling":[99,151],"precise":[102],"top-K":[106],"tokens.First,":[107],"we":[108,127,141],"propose":[109,128],"Windowed":[110],"Rotary":[111],"Position":[112],"Embedding,":[113],"which":[114],"decouples":[115],"positional":[117],"dependency":[118],"query":[120],"states":[123],"after":[124],"position":[125],"embedding.Then,":[126],"query-aware":[129],"that":[132,161],"optimizes":[133],"objective":[135],"score":[138],"directly.Finally,":[140],"design":[142],"heterogeneous":[144],"inference":[145],"architecture":[146],"offloading,":[150],"long":[152,182],"with":[155,171],"larger":[156],"batch":[157],"sizes.Experimental":[158],"results":[159],"demonstrate":[160],"achieve":[166],"lower":[168,174],"performance":[169],"similar":[172],"or":[173],"compared":[176],"existing":[178],"methods,":[179],"increasing":[181],"throughput":[185],"up":[187],"2.7.":[189]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}