{"id":"https://openalex.org/W7138345393","doi":"https://doi.org/10.1609/aaai.v40i25.39241","title":"Efficient Multimodal Large Language Model via Dynamic KV Cache Quantization","display_name":"Efficient Multimodal Large Language Model via Dynamic KV Cache Quantization","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138345393","doi":"https://doi.org/10.1609/aaai.v40i25.39241"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i25.39241","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i25.39241","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i25.39241","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129716481","display_name":"Jiahao Fan","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Jiahao Fan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5068618830","display_name":"Chien\u2010Ming Chen","orcid":"https://orcid.org/0000-0002-6502-472X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chien-Ming Chen","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5129716481"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.70618941,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"25","first_page":"20994","last_page":"21001"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9186000227928162,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9186000227928162,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.03150000050663948,"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"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.007000000216066837,"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/quantization","display_name":"Quantization (signal processing)","score":0.8299999833106995},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.6259999871253967},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.552299976348877},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.515500009059906},{"id":"https://openalex.org/keywords/cpu-cache","display_name":"CPU cache","score":0.43779999017715454},{"id":"https://openalex.org/keywords/computational-complexity-theory","display_name":"Computational complexity theory","score":0.30169999599456787}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8299999833106995},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8133999705314636},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.6259999871253967},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.552299976348877},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.515500009059906},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.43779999017715454},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.43549999594688416},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4081000089645386},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3968000113964081},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3407999873161316},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.30169999599456787},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.2973000109195709},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2736999988555908},{"id":"https://openalex.org/C5317259","wikidata":"https://www.wikidata.org/wiki/Q4462361","display_name":"Trellis quantization","level":5,"score":0.2711000144481659},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.2558000087738037},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.25380000472068787},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2500999867916107}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i25.39241","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i25.39241","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i25.39241","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i25.39241","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1],"language":[2],"models":[3],"(LMMs)":[4],"have":[5],"demonstrated":[6],"remarkable":[7],"capabilities":[8],"across":[9,138],"diverse":[10],"vision-language":[11],"tasks,":[12],"including":[13],"image":[14],"captioning,":[15],"visual":[16],"question":[17],"answering,":[18],"and":[19,26,71,91,136,159],"text-image":[20],"retrieval.":[21],"However,":[22],"their":[23,77],"computational":[24],"complexity":[25],"memory":[27,157],"footprint,":[28],"particularly":[29],"in":[30],"the":[31,107],"key-value":[32],"(KV)":[33],"cache":[34],"during":[35],"inference,":[36],"pose":[37],"significant":[38],"challenges":[39],"for":[40,61],"real-time":[41,102],"deployment,":[42],"especially":[43],"on":[44,101,133],"resource-constrained":[45],"devices.":[46],"In":[47],"this":[48],"paper,":[49],"we":[50,86,116],"propose":[51],"Dynamic":[52],"KV":[53,122],"Cache":[54],"Quantization,":[55],"a":[56,153],"novel":[57],"quantization":[58,68,73,98,150],"strategy":[59],"tailored":[60],"multimodal":[62,140],"LMMs.":[63],"Our":[64],"approach":[65],"applies":[66],"per-channel":[67],"to":[69,74,81],"(K)":[70],"per-token":[72],"(V),":[75],"leveraging":[76],"respective":[78],"statistical":[79],"distributions":[80],"optimize":[82],"precision":[83],"allocation.":[84],"Additionally,":[85],"introduce":[87],"an":[88],"adaptive":[89,129],"token":[90],"channel":[92],"recording":[93],"mechanism":[94],"that":[95,143],"dynamically":[96],"adjusts":[97],"parameters":[99],"based":[100],"distribution":[103],"tracking,":[104],"effectively":[105],"mitigating":[106],"impact":[108],"of":[109],"outliers.":[110],"To":[111],"further":[112],"enhance":[113],"compression":[114],"efficiency,":[115],"implement":[117],"fine-grained":[118],"grouping,":[119],"which":[120],"partitions":[121],"tensors":[123],"into":[124],"localized":[125],"subgroups,":[126],"enabling":[127],"more":[128],"quantization.":[130],"Experimental":[131],"results":[132],"LLaVA-1.5":[134],"(7B/13B)":[135],"Qwen-VL":[137],"multiple":[139],"benchmarks":[141],"demonstrate":[142],"our":[144],"method":[145],"significantly":[146],"outperforms":[147],"existing":[148],"KV-cache":[149],"approaches,":[151],"achieving":[152],"superior":[154],"trade-off":[155],"between":[156],"efficiency":[158],"model":[160],"accuracy.":[161]},"counts_by_year":[],"updated_date":"2026-03-18T06:31:55.123368","created_date":"2026-03-18T00:00:00"}