{"id":"https://openalex.org/W4412888967","doi":"https://doi.org/10.18653/v1/2025.findings-acl.33","title":"FlashBack: Efficient Retrieval-Augmented Language Modeling for Fast Inference","display_name":"FlashBack: Efficient Retrieval-Augmented Language Modeling for Fast Inference","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888967","doi":"https://doi.org/10.18653/v1/2025.findings-acl.33"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.33","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.33","pdf_url":"https://aclanthology.org/2025.findings-acl.33.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.33.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5008060076","display_name":"Runheng Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I125839683","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74","country_code":"CN","type":"education","lineage":["https://openalex.org/I125839683","https://openalex.org/I890469752"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Runheng Liu","raw_affiliation_strings":["School of Computer Science and Technology , Beijing Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Beijing Institute of Technology","institution_ids":["https://openalex.org/I125839683"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112683858","display_name":"Xingchen Xiao","orcid":null},"institutions":[{"id":"https://openalex.org/I125839683","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74","country_code":"CN","type":"education","lineage":["https://openalex.org/I125839683","https://openalex.org/I890469752"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xingchen Xiao","raw_affiliation_strings":["School of Computer Science and Technology , Beijing Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Beijing Institute of Technology","institution_ids":["https://openalex.org/I125839683"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087631670","display_name":"Heyan Huang","orcid":"https://orcid.org/0000-0002-0320-7520"},"institutions":[{"id":"https://openalex.org/I125839683","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74","country_code":"CN","type":"education","lineage":["https://openalex.org/I125839683","https://openalex.org/I890469752"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Heyan Huang","raw_affiliation_strings":["School of Computer Science and Technology , Beijing Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Beijing Institute of Technology","institution_ids":["https://openalex.org/I125839683"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077329331","display_name":"Zewen Chi","orcid":"https://orcid.org/0000-0003-1615-1885"},"institutions":[{"id":"https://openalex.org/I125839683","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74","country_code":"CN","type":"education","lineage":["https://openalex.org/I125839683","https://openalex.org/I890469752"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zewen Chi","raw_affiliation_strings":["School of Computer Science and Technology , Beijing Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Beijing Institute of Technology","institution_ids":["https://openalex.org/I125839683"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5002597610","display_name":"Zhijing Wu","orcid":"https://orcid.org/0000-0003-2473-3746"},"institutions":[{"id":"https://openalex.org/I125839683","display_name":"Beijing Institute of Technology","ror":"https://ror.org/01skt4w74","country_code":"CN","type":"education","lineage":["https://openalex.org/I125839683","https://openalex.org/I890469752"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhijing Wu","raw_affiliation_strings":["School of Computer Science and Technology , Beijing Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Beijing Institute of Technology","institution_ids":["https://openalex.org/I125839683"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I125839683"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"595","last_page":"608"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9638000130653381,"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":0.9638000130653381,"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":0.9397000074386597,"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/computer-science","display_name":"Computer science","score":0.7587258815765381},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7188176512718201},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.477582722902298},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4309074878692627},{"id":"https://openalex.org/keywords/information-retrieval","display_name":"Information retrieval","score":0.3582782745361328},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.3442091941833496}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7587258815765381},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7188176512718201},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.477582722902298},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4309074878692627},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.3582782745361328},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.3442091941833496}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.33","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.33","pdf_url":"https://aclanthology.org/2025.findings-acl.33.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"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.33","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.33","pdf_url":"https://aclanthology.org/2025.findings-acl.33.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":[{"display_name":"Quality Education","score":0.4099999964237213,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888967.pdf","grobid_xml":"https://content.openalex.org/works/W4412888967.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2055243143","https://openalex.org/W1986418932","https://openalex.org/W2357796999","https://openalex.org/W4321636575","https://openalex.org/W2741131631","https://openalex.org/W2045526782","https://openalex.org/W2156919374","https://openalex.org/W1483472507","https://openalex.org/W1984019423","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Retrieval-Augmented":[0],"Language":[1],"Modeling":[2],"(RALM)":[3],"by":[4,36,97],"integrating":[5],"large":[6],"language":[7,138],"models":[8,156],"(LLM)":[9],"with":[10,43,86,169],"relevant":[11],"documents":[12,102],"from":[13],"an":[14],"external":[15],"corpus":[16],"is":[17,150,175],"a":[18,38,51,75,151,185],"proven":[19],"methodology":[20],"for":[21,125],"enabling":[22],"the":[23,29,48,57,61,68,81,87,104,107,112,117,127,142,147,181,191],"LLM":[24,187],"to":[25,47,66,79,109,177],"generate":[26],"information":[27],"beyond":[28],"scope":[30],"of":[31,40,60,84,106],"its":[32],"pre-training":[33],"corpus.Previous":[34],"work":[35],"retrieving":[37],"set":[39],"tokens":[41,124],"iteratively":[42],"retrieved":[44,101],"content":[45],"prepending":[46,182],"input":[49],"poses":[50],"high":[52],"run-time":[53],"issue,":[54],"which":[55],"degrades":[56],"inference":[58,82,167,173],"efficiency":[59,83],"LLMs":[62],"because":[63],"they":[64],"fail":[65],"use":[67],"Key-Value":[69],"(KV)":[70],"cache":[71],"efficiently.We":[72],"propose":[73],"FLASHBACK,":[74],"modular":[76],"RALM":[77,85],"designed":[78],"improve":[80,137],"appending":[88,128],"context":[89,108,129,159,171],"pattern":[90],"while":[91],"maintaining":[92],"decent":[93],"performance":[94,140],"after":[95],"fine-tuning":[96,155],"Low-Rank":[98],"Adaption.FLASHBACK":[99],"appends":[100],"at":[103],"end":[105],"efficiently":[110],"utilize":[111],"KV":[113],"cache.We":[114],"also":[115],"introduce":[116],"Marking":[118,148],"Token":[119,149],"as":[120],"two":[121],"special":[122],"prompt":[123],"marking":[126],"during":[130],"fine-tuning.Our":[131],"experiments":[132],"show":[133],"that":[134,146],"FLASHBACK":[135,164],"can":[136],"modeling":[139],"in":[141,190],"perplexity":[143],"metric.We":[144],"proved":[145],"usable":[152],"add-on":[153],"when":[154],"on":[157,184],"specific":[158],"patterns.By":[160],"bypassing":[161],"unnecessary":[162],"recomputation,":[163],"achieves":[165],"fast":[166],"speed":[168,174],"long":[170],"input.The":[172],"up":[176],"4":[178],"faster":[179],"than":[180],"counterpart":[183],"7B":[186],"(Llama":[188],"2)":[189],"runtime":[192],"test.":[193]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}