{"id":"https://openalex.org/W7125280553","doi":"https://doi.org/10.1145/3786655","title":"From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation","display_name":"From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation","publication_year":2026,"publication_date":"2026-04-02","ids":{"openalex":"https://openalex.org/W7125280553","doi":"https://doi.org/10.1145/3786655"},"language":"en","primary_location":{"id":"doi:10.1145/3786655","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786655","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1145/3786655","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5123480256","display_name":"Jiahao Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiahao Wang","raw_affiliation_strings":["Hangzhou Dianzi University, Hangzhou, China and Approaching.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0001-8030-5976","affiliations":[{"raw_affiliation_string":"Hangzhou Dianzi University, Hangzhou, China and Approaching.AI, Beijing, China","institution_ids":["https://openalex.org/I50760025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123513924","display_name":"Weiyu Xie","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weiyu Xie","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-0173-1027","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123535843","display_name":"Mingxing Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mingxing Zhang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-7518-0753","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123484040","display_name":"Boxing Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Boxin Zhang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0004-2857-3713","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123527883","display_name":"Jianwei Dong","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jianwei Dong","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0007-3175-1607","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019728203","display_name":"Yuening Zhu","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yuening Zhu","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-6716-5709","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123502182","display_name":"Chen Lin","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen Lin","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0002-8554-0411","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jingqi Tang","orcid":"https://orcid.org/0009-0006-3300-9425"},"institutions":[{"id":"https://openalex.org/I4210131127","display_name":"Sound Approach (United Kingdom)","ror":"https://ror.org/02k9x1e74","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210131127"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Jingqi Tang","raw_affiliation_strings":["Approaching.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-3300-9425","affiliations":[{"raw_affiliation_string":"Approaching.AI, Beijing, China","institution_ids":["https://openalex.org/I4210131127"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080378174","display_name":"Yaochen Han","orcid":null},"institutions":[{"id":"https://openalex.org/I4210131127","display_name":"Sound Approach (United Kingdom)","ror":"https://ror.org/02k9x1e74","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210131127"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Yaochen Han","raw_affiliation_strings":["Approaching.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0004-7744-4707","affiliations":[{"raw_affiliation_string":"Approaching.AI, Beijing, China","institution_ids":["https://openalex.org/I4210131127"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123459143","display_name":"Zhiyuan Ai","orcid":null},"institutions":[{"id":"https://openalex.org/I4210131127","display_name":"Sound Approach (United Kingdom)","ror":"https://ror.org/02k9x1e74","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210131127"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Zhiyuan Ai","raw_affiliation_strings":["Approaching.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0007-8050-6288","affiliations":[{"raw_affiliation_string":"Approaching.AI, Beijing, China","institution_ids":["https://openalex.org/I4210131127"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123478284","display_name":"Xianglin Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I4210131127","display_name":"Sound Approach (United Kingdom)","ror":"https://ror.org/02k9x1e74","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210131127"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Xianglin Chen","raw_affiliation_strings":["Approaching.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-4975-4856","affiliations":[{"raw_affiliation_string":"Approaching.AI, Beijing, China","institution_ids":["https://openalex.org/I4210131127"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123478375","display_name":"Yongwei Wu","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yongwei Wu","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-6651-7032","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5036326577","display_name":"Congfeng Jiang","orcid":"https://orcid.org/0000-0003-3592-0328"},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Congfeng Jiang","raw_affiliation_strings":["Hangzhou Dianzi University, Hangzhou, China"],"raw_orcid":"https://orcid.org/0000-0003-3592-0328","affiliations":[{"raw_affiliation_string":"Hangzhou Dianzi University, Hangzhou, China","institution_ids":["https://openalex.org/I50760025"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.07657044,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"4","issue":"1","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":0.6215000152587891,"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":0.6215000152587891,"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":0.16779999434947968,"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/T13629","display_name":"Text Readability and Simplification","score":0.025200000032782555,"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/cache","display_name":"Cache","score":0.77920001745224},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.65420001745224},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5810999870300293},{"id":"https://openalex.org/keywords/preprocessor","display_name":"Preprocessor","score":0.5755000114440918},{"id":"https://openalex.org/keywords/prefix","display_name":"Prefix","score":0.5428000092506409},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.46709999442100525}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8309000134468079},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.77920001745224},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.65420001745224},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5810999870300293},{"id":"https://openalex.org/C34736171","wikidata":"https://www.wikidata.org/wiki/Q918333","display_name":"Preprocessor","level":2,"score":0.5755000114440918},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.5428000092506409},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.46709999442100525},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4478999972343445},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.34200000762939453},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.328900009393692},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.32710000872612},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.3021000027656555},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2815999984741211},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2705000042915344},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.26910001039505005},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26339998841285706},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3786655","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786655","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM on Management of Data","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:2601.12904","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.12904","pdf_url":"https://arxiv.org/pdf/2601.12904","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,"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"}],"best_oa_location":{"id":"doi:10.1145/3786655","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786655","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM on Management of Data","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G6224704988","display_name":null,"funder_award_id":"92467102","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Retrieval-Augmented":[0],"Generation":[1],"enhances":[2],"Large":[3],"Language":[4],"Models":[5],"by":[6,191],"integrating":[7],"external":[8],"knowledge,":[9],"which":[10],"reduces":[11,189],"hallucinations":[12],"but":[13],"increases":[14],"prompt":[15],"length.":[16],"This":[17],"increase":[18],"leads":[19,56],"to":[20,27,37,46,57,78,152,166,181,194],"higher":[21,183],"computational":[22],"costs":[23],"and":[24,100,149,188],"longer":[25],"Time":[26],"First":[28],"Token.":[29],"To":[30],"mitigate":[31],"this":[32],"issue,":[33],"existing":[34],"solutions":[35],"aim":[36],"reuse":[38,70,79],"the":[39,50,65,80,98,106,123,129,134,161,176],"preprocessed":[40],"KVCache":[41,69,82,130],"of":[42,52,68,103,175],"each":[43,119],"retrieved":[44],"chunk":[45,83],"accelerate":[47],"RAG.":[48,104],"However,":[49],"lack":[51],"cross-chunk":[53],"contextual":[54],"information":[55,112],"a":[58,91,139,143],"significant":[59],"drop":[60],"in":[61,76,122],"generation":[62,86,147,158],"quality,":[63],"leaving":[64],"potential":[66],"benefits":[67],"largely":[71],"unfulfilled.":[72],"The":[73],"challenge":[74],"lies":[75],"how":[77],"precomputed":[81],"while":[84,121],"preserving":[85],"quality.":[87],"We":[88],"propose":[89],"FusionRAG,":[90],"novel":[92],"inference":[93],"framework":[94],"that":[95,133],"optimizes":[96],"both":[97],"preprocessing":[99,108],"reprocessing":[101,125],"stages":[102],"In":[105],"offline":[107],"stage,":[109,126],"we":[110,127,141],"embed":[111],"from":[113],"other":[114],"related":[115],"text":[116],"chunks":[117],"into":[118],"chunk,":[120],"online":[124],"recompute":[128],"for":[131],"tokens":[132],"model":[135],"focuses":[136],"on.":[137],"As":[138],"result,":[140],"achieve":[142],"better":[144],"trade-off":[145],"between":[146],"quality":[148,159],"efficiency.":[150],"According":[151],"our":[153],"experiments,":[154],"FusionRAG":[155,178],"significantly":[156],"improves":[157],"at":[160],"same":[162],"recomputation":[163],"ratio":[164],"compared":[165,193],"previous":[167],"state-of-the-art":[168],"solutions.":[169],"By":[170],"recomputing":[171],"fewer":[172],"than":[173,186],"15%":[174],"tokens,":[177],"achieves":[179],"up":[180],"70%":[182],"normalized-F1":[184],"scores":[185],"baselines":[187],"TTFT":[190],"2.66-9.39\u00d7":[192],"Full":[195],"Attention.":[196]},"counts_by_year":[],"updated_date":"2026-08-05T07:39:15.569665","created_date":"2026-01-22T00:00:00"}