{"id":"https://openalex.org/W4411403197","doi":"https://doi.org/10.1145/3725273","title":"Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation","display_name":"Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation","publication_year":2025,"publication_date":"2025-06-17","ids":{"openalex":"https://openalex.org/W4411403197","doi":"https://doi.org/10.1145/3725273"},"language":"en","primary_location":{"id":"doi:10.1145/3725273","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3725273","pdf_url":null,"source":{"id":"https://openalex.org/S4387289859","display_name":"Proceedings of the ACM on Management of Data","issn_l":"2836-6573","issn":["2836-6573"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM on Management of Data","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100613382","display_name":"Shubham Agarwal","orcid":"https://orcid.org/0000-0003-3290-6328"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shubham Agarwal","raw_affiliation_strings":["Adobe Research, Bangalore, India"],"raw_orcid":"https://orcid.org/0000-0003-3290-6328","affiliations":[{"raw_affiliation_string":"Adobe Research, Bangalore, India","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Sai Sundaresan","orcid":"https://orcid.org/0009-0004-6908-6823"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sai Sundaresan","raw_affiliation_strings":["Adobe Research, Bangalore, India"],"raw_orcid":"https://orcid.org/0009-0004-6908-6823","affiliations":[{"raw_affiliation_string":"Adobe Research, Bangalore, India","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Subrata Mitra","orcid":"https://orcid.org/0009-0009-8436-3119"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Subrata Mitra","raw_affiliation_strings":["Adobe Research, Bangalore, India"],"raw_orcid":"https://orcid.org/0009-0009-8436-3119","affiliations":[{"raw_affiliation_string":"Adobe Research, Bangalore, India","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5057945543","display_name":"Debabrata Mahapatra","orcid":"https://orcid.org/0000-0001-7229-2944"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Debabrata Mahapatra","raw_affiliation_strings":["Adobe Research, Bangalore, India"],"raw_orcid":"https://orcid.org/0000-0001-7229-2944","affiliations":[{"raw_affiliation_string":"Adobe Research, Bangalore, India","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Archit Gupta","orcid":"https://orcid.org/0009-0008-2707-9958"},"institutions":[{"id":"https://openalex.org/I162827531","display_name":"Indian Institute of Technology Bombay","ror":"https://ror.org/02qyf5152","country_code":"IN","type":"education","lineage":["https://openalex.org/I162827531"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Archit Gupta","raw_affiliation_strings":["IIT Bombay, Mumbai, India"],"raw_orcid":"https://orcid.org/0009-0008-2707-9958","affiliations":[{"raw_affiliation_string":"IIT Bombay, Mumbai, India","institution_ids":["https://openalex.org/I162827531"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Rounak Sharma","orcid":"https://orcid.org/0009-0009-8205-040X"},"institutions":[{"id":"https://openalex.org/I94234084","display_name":"Indian Institute of Technology Kanpur","ror":"https://ror.org/05pjsgx75","country_code":"IN","type":"education","lineage":["https://openalex.org/I94234084"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Rounak Sharma","raw_affiliation_strings":["IIT Kanpur, Kanpur, India"],"raw_orcid":"https://orcid.org/0009-0009-8205-040X","affiliations":[{"raw_affiliation_string":"IIT Kanpur, Kanpur, India","institution_ids":["https://openalex.org/I94234084"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5118443880","display_name":"Nirmal Joshua Kapu","orcid":null},"institutions":[{"id":"https://openalex.org/I94234084","display_name":"Indian Institute of Technology Kanpur","ror":"https://ror.org/05pjsgx75","country_code":"IN","type":"education","lineage":["https://openalex.org/I94234084"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Nirmal Joshua Kapu","raw_affiliation_strings":["IIT Kanpur, Kanpur, India"],"raw_orcid":"https://orcid.org/0009-0008-8028-0386","affiliations":[{"raw_affiliation_string":"IIT Kanpur, Kanpur, India","institution_ids":["https://openalex.org/I94234084"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100767817","display_name":"Tong Yu","orcid":"https://orcid.org/0000-0002-5991-2050"},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tong Yu","raw_affiliation_strings":["Adobe Research, San Jose, USA"],"raw_orcid":"https://orcid.org/0000-0002-5991-2050","affiliations":[{"raw_affiliation_string":"Adobe Research, San Jose, USA","institution_ids":["https://openalex.org/I1306409833"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5055018529","display_name":"Shiv Kumar Saini","orcid":"https://orcid.org/0000-0001-6568-7104"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shiv Saini","raw_affiliation_strings":["Adobe Research, Bangalore, India"],"raw_orcid":"https://orcid.org/0000-0001-6568-7104","affiliations":[{"raw_affiliation_string":"Adobe Research, Bangalore, India","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":16.7519,"has_fulltext":false,"cited_by_count":12,"citation_normalized_percentile":{"value":0.99020788,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":97,"max":100},"biblio":{"volume":"3","issue":"3","first_page":"1","last_page":"28"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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":0.9998999834060669,"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":0.996999979019165,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.991599977016449,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.902553915977478},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.7685848474502563},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.7018737196922302},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.4200472831726074}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.902553915977478},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.7685848474502563},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.7018737196922302},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4200472831726074},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C18903297","wikidata":"https://www.wikidata.org/wiki/Q7150","display_name":"Ecology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3725273","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3725273","pdf_url":null,"source":{"id":"https://openalex.org/S4387289859","display_name":"Proceedings of the ACM on Management of Data","issn_l":"2836-6573","issn":["2836-6573"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM on Management of Data","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":40,"referenced_works":["https://openalex.org/W95608104","https://openalex.org/W1586825695","https://openalex.org/W2035642982","https://openalex.org/W2049009664","https://openalex.org/W2173213060","https://openalex.org/W2734941459","https://openalex.org/W2765113847","https://openalex.org/W2795530455","https://openalex.org/W2811266412","https://openalex.org/W2963929190","https://openalex.org/W2982157693","https://openalex.org/W2983309655","https://openalex.org/W3027879771","https://openalex.org/W3029685757","https://openalex.org/W3037785458","https://openalex.org/W3095488153","https://openalex.org/W3100965700","https://openalex.org/W3115947671","https://openalex.org/W3130554079","https://openalex.org/W3173814445","https://openalex.org/W3189877953","https://openalex.org/W3190126809","https://openalex.org/W3196447082","https://openalex.org/W4214686755","https://openalex.org/W4224903119","https://openalex.org/W4281758439","https://openalex.org/W4287124724","https://openalex.org/W4366341968","https://openalex.org/W4387321091","https://openalex.org/W4388584589","https://openalex.org/W4388979610","https://openalex.org/W4389315095","https://openalex.org/W4391376033","https://openalex.org/W4391876619","https://openalex.org/W4392867129","https://openalex.org/W4394892775","https://openalex.org/W4400525811","https://openalex.org/W4401043055","https://openalex.org/W4402727245","https://openalex.org/W4402778510"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W2384475851","https://openalex.org/W2000444236","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109"],"abstract_inverted_index":{"Retrieval-Augmented":[0],"Generation":[1],"(RAG)":[2],"is":[3,52],"often":[4],"used":[5],"with":[6,92,213],"Large":[7],"Language":[8],"Models":[9],"(LLMs)":[10],"to":[11,39,98,104,130,145,152,160,170,248],"infuse":[12],"domain":[13],"knowledge":[14,33],"or":[15,91],"user-specific":[16],"information.":[17],"In":[18,114],"RAG,":[19],"given":[20,50],"a":[21,24,32,121,155,217,223,237],"user":[22,56],"query,":[23],"retriever":[25],"extracts":[26],"chunks":[27,36,86,133],"of":[28,44,158],"relevant":[29],"text":[30,132],"from":[31],"base.":[34],"These":[35],"are":[37,149],"sent":[38],"an":[40],"LLM":[41],"as":[42,79,190,192],"part":[43],"the":[45,69,76,131,162,177,231],"input":[46,77],"prompt.":[47],"Typically,":[48],"any":[49,184],"chunk":[51],"repeatedly":[53,74],"retrieved":[54],"across":[55],"questions.":[57],"However,":[58],"currently,":[59],"for":[60,75,123,179,229],"every":[61],"question,":[62],"attention":[63],"layers":[64],"in":[65,139,176,227,242],"LLMs":[66],"fully":[67],"compute":[68],"Keys":[70],"and":[71,111,125,164,168,173,207,233,236,239],"Values":[72],"(KVs)":[73],"chunks,":[78],"state-of-the-art":[80],"methods":[81],"cannot":[82],"reuse":[83,96,181],"KV-caches":[84],"when":[85],"appear":[87],"at":[88],"arbitrary":[89,93],"locations":[90],"contexts.":[94],"Naive":[95],"leads":[97,103],"output":[99,166],"quality":[100],"degradation.":[101],"This":[102],"potentially":[105],"redundant":[106,200],"computations":[107],"on":[108,216],"expensive":[109],"GPUs":[110],"increases":[112],"latency.":[113],"this":[115],"work,":[116],"we":[117,135,195,221],"propose":[118],"Cache-Craft":[119,198],",":[120],"system":[122],"managing":[124],"reusing":[126],"precomputed":[127],"KVs":[128],"corresponding":[129],"(which":[134],"call":[136],"chunk-caches":[137,147,175],")":[138],"RAG-based":[140],"systems.":[141],"We":[142],"present":[143],"how":[144,151,169],"identify":[146],"that":[148,197],"reusable,":[150],"efficiently":[153,171],"perform":[154],"small":[156],"fraction":[157],"recomputation":[159],"fix":[161],"cache":[163],"maintain":[165],"quality,":[167],"store":[172],"evict":[174],"hardware":[178],"maximizing":[180],"while":[182,250],"masking":[183],"overheads.":[185],"With":[186],"real":[187,218],"production":[188,219],"workloads":[189],"well":[191],"synthetic":[193],"datasets,":[194],"show":[196],"reduces":[199],"computation":[201],"by":[202],"51%":[203],"over":[204,209],"SOTA":[205],"prefix-caching":[206],"75%":[208],"full":[210],"recomputation.":[211],"Additionally,":[212],"continuous":[214],"batching":[215],"workload,":[220],"get":[222],"1.6\u00d7":[224],"speed":[225],"up":[226],"throughput":[228],"both":[230],"LLama-3-8B":[232],"70B":[234],"models":[235],"2.1\u00d7":[238],"2\u00d7":[240],"reduction":[241],"end-to-end":[243],"response":[244],"latency":[245],"respectively,":[246],"compared":[247],"prefix-caching,":[249],"maintaining":[251],"generation":[252],"quality.":[253]},"counts_by_year":[{"year":2026,"cited_by_count":8},{"year":2025,"cited_by_count":4}],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}