{"id":"https://openalex.org/W7214770294","doi":"https://doi.org/10.48550/arxiv.2609.33503","title":"RelaxKV: Recomputation Guided by the Query with Sparse Context Attention for Efficient KV Cache Reuse","display_name":"RelaxKV: Recomputation Guided by the Query with Sparse Context Attention for Efficient KV Cache Reuse","publication_year":2026,"publication_date":"2026-09-27","ids":{"openalex":"https://openalex.org/W7214770294","doi":"https://doi.org/10.48550/arxiv.2609.33503"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.33503","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33503","pdf_url":"https://arxiv.org/pdf/2609.33503","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.33503","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124085287","display_name":"Ruoling Qi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]},{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qi, Ruoling","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom","Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]},{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153460949","display_name":"Yirui Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liu, Yirui","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147124197","display_name":"Xuaner Wu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wu, Xuaner","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147112299","display_name":"Yuxin Jin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jin, Yuxin","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153512122","display_name":"Jian Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I63190737","display_name":"University at Buffalo, State University of New York","ror":"https://ror.org/01y64my43","country_code":"US","type":"education","lineage":["https://openalex.org/I63190737"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chen, Jian","raw_affiliation_strings":["State University of New York at Buffalo"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"State University of New York at Buffalo","institution_ids":["https://openalex.org/I63190737"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101392776","display_name":"Jiayu Qin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I63190737","display_name":"University at Buffalo, State University of New York","ror":"https://ror.org/01y64my43","country_code":"US","type":"education","lineage":["https://openalex.org/I63190737"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Qin, Jiayu","raw_affiliation_strings":["State University of New York at Buffalo"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"State University of New York at Buffalo","institution_ids":["https://openalex.org/I63190737"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153488952","display_name":"Yin Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Yin","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5102806605","display_name":"Jiawei Shao","orcid":"https://orcid.org/0000-0001-8836-1430","observed_orcids":["https://orcid.org/0000-0001-8836-1430"]},"institutions":[{"id":"https://openalex.org/I4210136246","display_name":"China Telecom (China)","ror":"https://ror.org/03jgnzt20","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210136246"]},{"id":"https://openalex.org/I4387153335","display_name":"China Telecom","ror":"https://ror.org/05p67dv18","country_code":"CN","type":"company","lineage":["https://openalex.org/I4387153335"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shao, Jiawei","raw_affiliation_strings":["Institute of Artificial Intelligence , China Telecom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Artificial Intelligence , China Telecom","institution_ids":["https://openalex.org/I4210136246","https://openalex.org/I4387153335"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10028","display_name":"Topic Modeling","score":0.7060999870300293,"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/T10028","display_name":"Topic Modeling","score":0.7060999870300293,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.22169999778270721,"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.012500000186264515,"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/kv-cache-reuse","display_name":"KV cache reuse","score":0.8029999732971191},{"id":"https://openalex.org/keywords/retrieval-augmented-generation","display_name":"retrieval-augmented generation","score":0.8840000033378601},{"id":"https://openalex.org/keywords/long-context-decoding","display_name":"long-context decoding","score":0.5669999718666077}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8773000240325928},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.7752000093460083},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.7336000204086304},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6528000235557556},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5825999975204468},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.4706000089645386},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.46779999136924744},{"id":"https://openalex.org/C2776036281","wikidata":"https://www.wikidata.org/wiki/Q48769818","display_name":"Constraint (computer-aided design)","level":2,"score":0.4343999922275543},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.384799987077713},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.365200012922287},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.3617999851703644},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.3244999945163727},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3224000036716461},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.32030001282691956},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.3140999972820282},{"id":"https://openalex.org/C5165142","wikidata":"https://www.wikidata.org/wiki/Q5432732","display_name":"False sharing","level":5,"score":0.29989999532699585},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.2992999851703644},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.2962000072002411},{"id":"https://openalex.org/C167713795","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"Smart Cache","level":5,"score":0.28600001335144043},{"id":"https://openalex.org/C113166858","wikidata":"https://www.wikidata.org/wiki/Q5015981","display_name":"Cache pollution","level":5,"score":0.2840999960899353}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.33503","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33503","pdf_url":"https://arxiv.org/pdf/2609.33503","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.33503","doi":"https://doi.org/10.48550/arxiv.2609.33503","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.33503","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.33503","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33503","pdf_url":"https://arxiv.org/pdf/2609.33503","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214770294.pdf","grobid_xml":"https://content.openalex.org/works/W7214770294.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7127607560","https://openalex.org/W7214786013","https://openalex.org/W7221171401","https://openalex.org/W7134058334","https://openalex.org/W7221781624","https://openalex.org/W7154612387","https://openalex.org/W7214979260","https://openalex.org/W7220409715","https://openalex.org/W7214245965","https://openalex.org/W7220726744"],"abstract_inverted_index":{"Cross-request":[0],"KV":[1,33],"caching":[2,14],"reduces":[3],"the":[4,52,63,67,92,147,165],"prefill":[5],"cost":[6],"of":[7,167],"Retrieval-Augmented":[8],"Generation":[9],"(RAG),":[10],"but":[11,31,49],"conventional":[12],"prefix":[13],"severely":[15],"limits":[16],"cache":[17,77],"reuse":[18],"across":[19,140],"requests.":[20],"Position-Independent":[21],"Caching":[22],"(PIC)":[23],"removes":[24],"this":[25],"constraint":[26],"by":[27,91],"reusing":[28],"independent":[29],"chunks,":[30],"their":[32,102],"states":[34,43,58],"miss":[35],"cross-chunk":[36],"interactions.":[37],"Existing":[38],"methods":[39],"selectively":[40],"recompute":[41],"token":[42],"to":[44,55,59,66,104],"recover":[45],"these":[46],"missing":[47],"interactions,":[48],"primarily":[50],"allocate":[51],"recomputation":[53,64,88,103,168],"budget":[54],"selecting":[56],"which":[57,74],"recompute,":[60],"while":[61],"fixing":[62],"context":[65,159,169],"full":[68],"causal":[69],"prefix.":[70],"We":[71],"introduce":[72],"RelaxKV,":[73],"formulates":[75],"selective":[76,149],"repair":[78,85,98],"as":[79],"a":[80,105,117,134,141],"joint":[81],"allocation":[82],"problem":[83],"over":[84,125],"targets":[86,99],"and":[87,100,145,153,157],"context.":[89],"Guided":[90],"user":[93],"query,":[94],"RelaxKV":[95,115,132],"identifies":[96],"layer-specific":[97],"restricts":[101],"query-relevant":[106],"context,":[107],"reducing":[108],"attention":[109],"computation.":[110],"Across":[111],"four":[112],"decoder":[113],"models,":[114],"at":[116,155],"15%":[118],"anchor":[119],"ratio":[120],"improves":[121],"aggregate":[122],"LongBench":[123],"performance":[124],"ProphetKV":[126,139],"on":[127,151],"all":[128],"models.":[129],"On":[130],"Qwen3-14B,":[131],"provides":[133],"stronger":[135],"quality-TTFT":[136],"trade-off":[137],"than":[138],"5%-30%":[142],"anchor-ratio":[143],"sweep,":[144],"achieves":[146],"best":[148],"results":[150],"RULER-MV":[152],"LV-Eval":[154],"16K":[156],"32K":[158],"lengths.":[160],"Controlled":[161],"ablations":[162],"further":[163],"demonstrate":[164],"importance":[166],"selection.":[170]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-30T00:00:00"}