{"id":"https://openalex.org/W7203664466","doi":"https://doi.org/10.48550/arxiv.2608.14555","title":"Discovering KV Cache Eviction Policies via LLM-Guided Program Evolution","display_name":"Discovering KV Cache Eviction Policies via LLM-Guided Program Evolution","publication_year":2026,"publication_date":"2026-05-12","ids":{"openalex":"https://openalex.org/W7203664466","doi":"https://doi.org/10.48550/arxiv.2608.14555"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.14555","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.14555","pdf_url":"https://arxiv.org/pdf/2608.14555","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.14555","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5025149802","display_name":"Pratik Poudel","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19700959","display_name":"Florida International University","ror":"https://ror.org/02gz6gg07","country_code":"US","type":"education","lineage":["https://openalex.org/I19700959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Poudel, Pratik","raw_affiliation_strings":["FIU Miami , FL , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"FIU Miami , FL , USA","institution_ids":["https://openalex.org/I19700959"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060093535","display_name":"Yanzhao Wu","orcid":"https://orcid.org/0000-0001-8761-5486","observed_orcids":["https://orcid.org/0000-0001-8761-5486"]},"institutions":[{"id":"https://openalex.org/I19700959","display_name":"Florida International University","ror":"https://ror.org/02gz6gg07","country_code":"US","type":"education","lineage":["https://openalex.org/I19700959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wu, Yanzhao","raw_affiliation_strings":["FIU Miami , FL , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"FIU Miami , FL , USA","institution_ids":["https://openalex.org/I19700959"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147391583","display_name":"Sumit Jha","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19700959","display_name":"Florida International University","ror":"https://ror.org/02gz6gg07","country_code":"US","type":"education","lineage":["https://openalex.org/I19700959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jha, Sumit","raw_affiliation_strings":["FIU Gainesville , FL , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"FIU Gainesville , FL , USA","institution_ids":["https://openalex.org/I19700959"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5083193232","display_name":"Jason Liu","orcid":"https://orcid.org/0000-0001-8222-4013","observed_orcids":["https://orcid.org/0000-0001-8222-4013","https://orcid.org/0009-0008-3689-986X"]},"institutions":[{"id":"https://openalex.org/I19700959","display_name":"Florida International University","ror":"https://ror.org/02gz6gg07","country_code":"US","type":"education","lineage":["https://openalex.org/I19700959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu, Jason","raw_affiliation_strings":["FIU Miami , FL , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"FIU Miami , FL , USA","institution_ids":["https://openalex.org/I19700959"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I19700959"],"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/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.3163999915122986,"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/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.3163999915122986,"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/T10260","display_name":"Software Engineering Research","score":0.16380000114440918,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.14010000228881836,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/kv-cache-compression","display_name":"KV cache compression","score":0.9789999723434448},{"id":"https://openalex.org/keywords/long-context-inference","display_name":"long-context inference","score":0.8830000162124634},{"id":"https://openalex.org/keywords/software-evolution","display_name":"software evolution","score":0.7540000081062317},{"id":"https://openalex.org/keywords/reward-hacking","display_name":"reward hacking","score":0.6259999871253967}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7752000093460083},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.7462000250816345},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.5041000247001648},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.4896000027656555},{"id":"https://openalex.org/C2778227907","wikidata":"https://www.wikidata.org/wiki/Q1893186","display_name":"Eviction","level":2,"score":0.4745999872684479},{"id":"https://openalex.org/C108154423","wikidata":"https://www.wikidata.org/wiki/Q1469792","display_name":"Salience (neuroscience)","level":2,"score":0.47380000352859497},{"id":"https://openalex.org/C165743212","wikidata":"https://www.wikidata.org/wiki/Q104555","display_name":"Ruler","level":2,"score":0.3995000123977661},{"id":"https://openalex.org/C187691185","wikidata":"https://www.wikidata.org/wiki/Q2020720","display_name":"Grid","level":2,"score":0.3917999863624573},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.37139999866485596},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.34040001034736633},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.30469998717308044},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.30230000615119934},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2773999869823456},{"id":"https://openalex.org/C132943942","wikidata":"https://www.wikidata.org/wiki/Q2562511","display_name":"Footprint","level":2,"score":0.27399998903274536},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.2660999894142151},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.26269999146461487},{"id":"https://openalex.org/C162319229","wikidata":"https://www.wikidata.org/wiki/Q175263","display_name":"Data structure","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.25099998712539673},{"id":"https://openalex.org/C180591934","wikidata":"https://www.wikidata.org/wiki/Q1253369","display_name":"Downtime","level":2,"score":0.250900000333786},{"id":"https://openalex.org/C110875604","wikidata":"https://www.wikidata.org/wiki/Q75","display_name":"The Internet","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.14555","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.14555","pdf_url":"https://arxiv.org/pdf/2608.14555","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.14555","doi":"https://doi.org/10.48550/arxiv.2608.14555","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.14555","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:2608.14555","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.14555","pdf_url":"https://arxiv.org/pdf/2608.14555","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7203664466.pdf","grobid_xml":"https://content.openalex.org/works/W7203664466.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W758372786","https://openalex.org/W4387294617","https://openalex.org/W4394778349","https://openalex.org/W4399414358","https://openalex.org/W4401306886","https://openalex.org/W4404609179","https://openalex.org/W4406071672","https://openalex.org/W4407123430","https://openalex.org/W4415108068","https://openalex.org/W4416035932","https://openalex.org/W6910722644","https://openalex.org/W7084400772","https://openalex.org/W7128714544","https://openalex.org/W7135011577"],"related_works":["https://openalex.org/W4416035932","https://openalex.org/W7162305397","https://openalex.org/W6979369520","https://openalex.org/W7116107563","https://openalex.org/W7172398105","https://openalex.org/W7220691001","https://openalex.org/W7167591330","https://openalex.org/W7128714544","https://openalex.org/W7125801894","https://openalex.org/W7160820266"],"abstract_inverted_index":{"KV":[0,45],"cache":[1,46],"compression":[2,34],"is":[3,133],"critical":[4],"for":[5,42,160,188],"long-context":[6],"inference,":[7],"yet":[8],"effective":[9],"eviction":[10,47],"policies":[11,48],"remain":[12],"difficult":[13],"to":[14],"design:":[15],"existing":[16],"prefill-stage":[17],"methods":[18],"often":[19],"rely":[20],"on":[21,112,117],"hand-crafted":[22],"salience":[23,74],"heuristics":[24],"that":[25,63,126,179],"can":[26],"be":[27],"brittle":[28],"across":[29,100],"models,":[30],"context":[31],"lengths,":[32],"and":[33,71,102,114,175,183],"ratios.":[35],"We":[36],"present":[37],"CacheCraft,":[38],"a":[39,59,157,164,168,176],"program-evolution":[40],"methodology":[41],"automatically":[43],"discovering":[44],"using":[49],"an":[50],"LLM-guided":[51],"code-evolution":[52],"engine.":[53],"CacheCraft":[54,155],"discovers":[55],"FRC":[56,82,153],"(Feature-Rich":[57],"Compression),":[58],"fixed-weight":[60],"three-signal":[61],"scorer":[62,140],"combines":[64],"local":[65],"attention":[66,69],"received,":[67],"neighborhood":[68],"density,":[70],"KV-head":[72],"maximum":[73],"with":[75,96,171],"chunk-level":[76],"top-k":[77],"selection.":[78],"Without":[79],"per-model":[80],"retuning,":[81],"ranks":[83],"first":[84],"among":[85],"the":[86,127,134,139,190],"evaluated":[87],"single-pass":[88],"KVPress":[89],"baselines":[90],"at":[91,119],"every":[92],"RULER":[93,143],"4k/8k":[94],"cell":[95],"r":[97],">=":[98],"0.75":[99],"Llama-3.1-8B-Instruct":[101],"Qwen3-8B":[103],"(12":[104],"of":[105],"20":[106],"grid":[107],"cells),":[108],"gaining":[109],"+15.4":[110],"points":[111,116],"Llama-4k":[113],"+13.9":[115],"Qwen-8k":[118],"88%":[120],"compression.":[121],"A":[122],"scorer-versus-structure":[123],"decomposition":[124],"shows":[125],"scoring":[128],"family,":[129],"not":[130],"chunk":[131,147],"selection,":[132],"load-bearing":[135],"design":[136],"choice:":[137],"incorporating":[138],"contributes":[141,149],"+67.2":[142],"points,":[144],"while":[145],"improving":[146],"structure":[148],"only":[150],"~0.1.":[151],"Beyond":[152],"itself,":[154],"provides":[156],"transferable":[158],"recipe":[159],"automated":[161],"eviction-policy":[162],"discovery:":[163],"compact":[165],"policy":[166],"interface,":[167],"cascade":[169],"evaluator":[170],"strict":[172],"output":[173],"invariants,":[174],"diagnostic":[177],"loop":[178],"treats":[180],"search":[181],"plateaus":[182],"reward-hacking":[184],"failures":[185],"as":[186],"evidence":[187],"reformulating":[189],"editable":[191],"interface.":[192]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-19T00:00:00"}
