{"id":"https://openalex.org/W4415708409","doi":"https://doi.org/10.1109/icme59968.2025.11209367","title":"AKVQ-VL: Attention-Aware KV Cache Adaptive 2-Bit Quantization for Vision-Language Models","display_name":"AKVQ-VL: Attention-Aware KV Cache Adaptive 2-Bit Quantization for Vision-Language Models","publication_year":2025,"publication_date":"2025-06-30","ids":{"openalex":"https://openalex.org/W4415708409","doi":"https://doi.org/10.1109/icme59968.2025.11209367"},"language":null,"primary_location":{"id":"doi:10.1109/icme59968.2025.11209367","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icme59968.2025.11209367","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Multimedia and Expo (ICME)","raw_type":"proceedings-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/A5086331195","display_name":"Zhiyong Su","orcid":"https://orcid.org/0000-0001-9483-5268"},"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":"Zunhai Su","raw_affiliation_strings":["Tsinghua University,Shenzhen International Graduate School,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University,Shenzhen International Graduate School,Shenzhen,China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Wang Shen","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang Shen","raw_affiliation_strings":["Huawei Technologies Co., Ltd.,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd.,Shenzhen,China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5117117476","display_name":"Linge Li","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Linge Li","raw_affiliation_strings":["Huawei Technologies Co., Ltd.,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd.,Shenzhen,China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100656819","display_name":"Zhe Chen","orcid":"https://orcid.org/0000-0002-3215-2696"},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhe Chen","raw_affiliation_strings":["Huawei Technologies Co., Ltd.,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd.,Shenzhen,China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116053899","display_name":"Hanyu Wei","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":"Hanyu Wei","raw_affiliation_strings":["Tsinghua University,Shenzhen International Graduate School,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University,Shenzhen International Graduate School,Shenzhen,China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056191914","display_name":"Huangqi Yu","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huangqi Yu","raw_affiliation_strings":["Huawei Technologies Co., Ltd.,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd.,Shenzhen,China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5102309342","display_name":"Kehong Yuan","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":"Kehong Yuan","raw_affiliation_strings":["Tsinghua University,Shenzhen International Graduate School,Shenzhen,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University,Shenzhen International Graduate School,Shenzhen,China","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.7655,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.76641896,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"6"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.7674999833106995,"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":0.7674999833106995,"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":0.09359999746084213,"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":0.015799999237060547,"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":0.8450000286102295},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.47620001435279846},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.47519999742507935},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.3569999933242798},{"id":"https://openalex.org/keywords/energy-consumption","display_name":"Energy consumption","score":0.31709998846054077}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8450000286102295},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7526999711990356},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.47620001435279846},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.47519999742507935},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.41029998660087585},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3962000012397766},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.3569999933242798},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.31709998846054077},{"id":"https://openalex.org/C2221639","wikidata":"https://www.wikidata.org/wiki/Q2877","display_name":"Discrete cosine transform","level":3,"score":0.31360000371932983},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.3046000003814697},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.28870001435279846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icme59968.2025.11209367","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icme59968.2025.11209367","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Multimedia and Expo (ICME)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W1779006005","https://openalex.org/W3168463823","https://openalex.org/W4388979610","https://openalex.org/W4392172801","https://openalex.org/W4402683788","https://openalex.org/W4402727764","https://openalex.org/W4404782265"],"related_works":[],"abstract_inverted_index":{"Vision-language":[0],"models":[1],"(VLMs)":[2],"show":[3],"remarkable":[4],"performance":[5],"in":[6,21,51,64,95],"multimodal":[7,12,48,121],"tasks.":[8],"However,":[9],"excessively":[10],"long":[11],"inputs":[13],"lead":[14],"to":[15,81,104,144],"oversized":[16],"Key-Value":[17],"(KV)":[18],"caches,":[19,108],"resulting":[20,50],"significant":[22],"memory":[23,138],"consumption":[24],"and":[25,66,76,120,149],"I/O":[26],"bottlenecks.":[27],"Previous":[28],"KV":[29,96,107],"quantization":[30,90,111,116],"methods":[31],"for":[32],"Large":[33],"Language":[34],"Models":[35],"(LLMs)":[36],"may":[37],"alleviate":[38],"these":[39],"issues":[40],"but":[41],"overlook":[42],"the":[43,59,71,100],"attention":[44],"saliency":[45,62],"differences":[46],"of":[47,114],"tokens,":[49],"suboptimal":[52],"performance.":[53],"In":[54],"this":[55],"paper,":[56],"we":[57],"investigate":[58],"attention-aware":[60],"token":[61],"patterns":[63,80],"VLM":[65],"propose":[67],"AKVQ-VL.":[68],"AKVQ-VL":[69,98,125,134],"leverages":[70],"proposed":[72],"Text-Salient":[73],"Attention":[74,78],"(TSA)":[75],"Pivot-Token-Salient":[77],"(PSA)":[79],"adaptively":[82],"allocate":[83],"bit":[84],"budgets.":[85],"Moreover,":[86],"achieving":[87],"extremely":[88],"low-bit":[89],"requires":[91],"effectively":[92],"addressing":[93],"outliers":[94],"tensors.":[97],"utilizes":[99],"Walsh-Hadamard":[101],"transform":[102],"(WHT)":[103],"construct":[105],"outlier-free":[106],"thereby":[109],"reducing":[110],"difficulty.":[112],"Evaluations":[113],"2-bit":[115],"on":[117],"12":[118],"long-context":[119],"tasks":[122],"demonstrate":[123],"that":[124],"maintains":[126],"or":[127],"even":[128],"improves":[129],"accuracy,":[130],"outperforming":[131],"LLM-oriented":[132],"methods.":[133],"can":[135],"reduce":[136],"peak":[137],"usage":[139],"by":[140],"2.13\u00d7,":[141],"support":[142],"up":[143],"3.25\u00d7":[145],"larger":[146],"batch":[147],"sizes":[148],"2.46\u00d7":[150],"throughput.":[151]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-30T00:00:00"}