{"id":"https://openalex.org/W4415538315","doi":"https://doi.org/10.1145/3746027.3755305","title":"BOLT: Fewer Tokens but More Performance Retention for Efficient Vision-Language Models Inference","display_name":"BOLT: Fewer Tokens but More Performance Retention for Efficient Vision-Language Models Inference","publication_year":2025,"publication_date":"2025-10-25","ids":{"openalex":"https://openalex.org/W4415538315","doi":"https://doi.org/10.1145/3746027.3755305"},"language":null,"primary_location":{"id":"doi:10.1145/3746027.3755305","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3746027.3755305","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 33rd ACM International Conference on Multimedia","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/A5109676444","display_name":"Jiahua Bao","orcid":"https://orcid.org/0000-0003-0610-4321"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jiahua Bao","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, China, and China Mobile 5G Institute, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, China, and China Mobile 5G Institute, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101603499","display_name":"Siyao Cheng","orcid":"https://orcid.org/0000-0002-1263-9907"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Siyao Cheng","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, China, and China Mobile 5G Institute, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, China, and China Mobile 5G Institute, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113871275","display_name":"Jiaxing Du","orcid":"https://orcid.org/0009-0003-6451-0181"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiaxing Du","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103203226","display_name":"Changjiang He","orcid":"https://orcid.org/0000-0003-4447-1130"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Changjiang He","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Zeming Lang","orcid":"https://orcid.org/0009-0007-1154-3157"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeming Lang","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100396946","display_name":"Hao Zhang","orcid":"https://orcid.org/0000-0002-6769-2115"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hao Zhang","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100454005","display_name":"Jie Liu","orcid":"https://orcid.org/0000-0001-6209-6886"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jie Liu","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, Heilongjiang, China, and China Mobile 5G Institute, Harbin, China"],"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China, National Key Laboratory of SFTaS, Harbin, Heilongjiang, China, and China Mobile 5G Institute, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5109676444"],"corresponding_institution_ids":["https://openalex.org/I204983213"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.29748835,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4058","last_page":"4067"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9998999834060669,"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":0.9998999834060669,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9988999962806702,"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":0.9962999820709229,"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/security-token","display_name":"Security token","score":0.8901000022888184},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7526999711990356},{"id":"https://openalex.org/keywords/merge","display_name":"Merge (version control)","score":0.7200000286102295},{"id":"https://openalex.org/keywords/categorization","display_name":"Categorization","score":0.5378000140190125},{"id":"https://openalex.org/keywords/proxy","display_name":"Proxy (statistics)","score":0.5263000130653381},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.3668999969959259}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8901000022888184},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8133000135421753},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7526999711990356},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.7200000286102295},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.5378000140190125},{"id":"https://openalex.org/C2780148112","wikidata":"https://www.wikidata.org/wiki/Q1432581","display_name":"Proxy (statistics)","level":2,"score":0.5263000130653381},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48730000853538513},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.3668999969959259},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3529999852180481},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3059000074863434},{"id":"https://openalex.org/C74296488","wikidata":"https://www.wikidata.org/wiki/Q2527392","display_name":"End-to-end principle","level":2,"score":0.27970001101493835},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.27549999952316284},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.2694999873638153},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2615000009536743},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.260699987411499},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.2587999999523163}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3746027.3755305","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3746027.3755305","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 33rd ACM International Conference on Multimedia","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G8766406793","display_name":null,"funder_award_id":"62350710797, 61972114, 62106061","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":10,"referenced_works":["https://openalex.org/W2085753599","https://openalex.org/W2307512708","https://openalex.org/W2963518342","https://openalex.org/W2979382951","https://openalex.org/W3004268082","https://openalex.org/W4387398228","https://openalex.org/W4389523832","https://openalex.org/W4394862623","https://openalex.org/W4402726948","https://openalex.org/W4404233915"],"related_works":[],"abstract_inverted_index":{"Vision-Language":[0],"Models":[1],"(VLMs)":[2],"have":[3],"achieved":[4],"significant":[5],"advances":[6],"across":[7],"various":[8],"downstream":[9],"tasks.":[10],"However,":[11],"as":[12],"their":[13],"performance":[14],"improves,":[15],"the":[16,68,106,110,115,122],"increasing":[17],"number":[18,44],"of":[19,45,70],"parameters":[20],"results":[21,141],"in":[22,74,124,159,166],"slower":[23],"prefilling":[24],"speeds":[25],"and":[26,58,93,104,127,135,138,162],"longer":[27],"inference":[28,167],"times.":[29],"To":[30,119],"overcome":[31],"these":[32],"limitations,":[33],"we":[34,50,102,130],"observe":[35],"that":[36,143],"most":[37,116],"VLMs":[38],"do":[39],"not":[40],"require":[41],"a":[42,56,149,156,163],"large":[43],"image":[46],"tokens":[47,85,108,112],"for":[48,121],"inference,":[49],"propose":[51],"BOLT":[52,66,132,144],"(Basis-Oriented":[53],"Lightweight":[54],"Token-Trimming),":[55],"training-free":[57],"cross-attention-free":[59],"token":[60,79,151],"compression":[61,152],"method.":[62],"Unlike":[63],"existing":[64],"approaches,":[65],"addresses":[67],"challenge":[69],"insufficient":[71],"visual":[72],"cues":[73],"textual":[75],"prompts":[76],"by":[77,97],"leveraging":[78],"internal":[80],"data":[81],"distributions.":[82],"We":[83],"categorize":[84],"into":[86],"three":[87],"types:":[88],"key":[89],"tokens,":[90,92],"proxy":[91,111],"remaining":[94,107],"tokens.":[95],"Then,":[96],"applying":[98],"basis":[99],"space":[100],"similarity,":[101],"merge":[103],"filter":[105],"with":[109,148],"to":[113,155],"retain":[114],"informative":[117],"ones.":[118],"account":[120],"differences":[123],"VLM":[125],"architectures":[126],"model":[128],"sizes,":[129],"evaluate":[131],"on":[133],"LLaVA-Next-Llama3":[134],"LLaVA-1.5":[136],"(7B":[137],"13B).":[139],"Our":[140],"show":[142],"achieves":[145],"state-of-the-art":[146],"performance,":[147],"90%":[150],"ratio":[153],"leading":[154],"3.3\u00d7":[157],"increase":[158],"pre-filling":[160],"speed":[161],"1.5\u00d7":[164],"improvement":[165],"speed,":[168],"outperforming":[169],"other":[170],"methods.":[171]},"counts_by_year":[],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-10-25T00:00:00"}