{"id":"https://openalex.org/W4417066054","doi":"https://doi.org/10.1109/iccv51701.2025.01986","title":"Skip-Vision: Efficient and Scalable Acceleration of Vision-Language Models Via Adaptive Token Skipping","display_name":"Skip-Vision: Efficient and Scalable Acceleration of Vision-Language Models Via Adaptive Token Skipping","publication_year":2025,"publication_date":"2025-10-19","ids":{"openalex":"https://openalex.org/W4417066054","doi":"https://doi.org/10.1109/iccv51701.2025.01986"},"language":"en","primary_location":{"id":"doi:10.1109/iccv51701.2025.01986","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.01986","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF International Conference on Computer Vision (ICCV)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2503.21817","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111275323","display_name":"Weili Zeng","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weili Zeng","raw_affiliation_strings":["AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014742942","display_name":"Ziyuan Huang","orcid":"https://orcid.org/0000-0002-8939-6456"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ziyuan Huang","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042803376","display_name":"Kaixiang Ji","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kaixiang Ji","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5102136293","display_name":"Y. Yan","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yichao Yan","raw_affiliation_strings":["AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"21384","last_page":"21397"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5275999903678894,"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":0.5275999903678894,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.1858000010251999,"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":0.07569999992847443,"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/scalability","display_name":"Scalability","score":0.6840000152587891},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6589999794960022},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.6517000198364258},{"id":"https://openalex.org/keywords/subnetwork","display_name":"Subnetwork","score":0.6468999981880188},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.6021999716758728},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.499099999666214},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4812000095844269},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.4666000008583069}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8194000124931335},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.6840000152587891},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6589999794960022},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6517000198364258},{"id":"https://openalex.org/C2780186347","wikidata":"https://www.wikidata.org/wiki/Q11414","display_name":"Subnetwork","level":2,"score":0.6468999981880188},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6021999716758728},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.499099999666214},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4812000095844269},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4666000008583069},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4636000096797943},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.45730000734329224},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44780001044273376},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.41440001130104065},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.3792000114917755},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.36340001225471497},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.32919999957084656},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.3098999857902527},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2928999960422516},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2921000123023987},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.28859999775886536},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2872999906539917},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2734000086784363},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.2720000147819519}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1109/iccv51701.2025.01986","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.01986","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF International Conference on Computer Vision (ICCV)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2503.21817","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.21817","pdf_url":"https://arxiv.org/pdf/2503.21817","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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"},{"id":"doi:10.48550/arxiv.2503.21817","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2503.21817","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2503.21817","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.21817","pdf_url":"https://arxiv.org/pdf/2503.21817","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G7308652284","display_name":null,"funder_award_id":"62201342","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":{"Transformer-based":[0],"models":[1],"have":[2],"driven":[3],"significant":[4],"advancements":[5],"in":[6,53],"Multimodal":[7],"Large":[8],"Language":[9],"Models":[10],"(MLLMs),":[11],"yet":[12],"their":[13],"computational":[14],"costs":[15],"surge":[16],"drastically":[17],"when":[18],"scaling":[19,158],"resolution,":[20],"training":[21,49,71,128],"data,":[22],"and":[23,50,138],"model":[24,120],"parameters.":[25],"A":[26],"key":[27],"bottleneck":[28],"stems":[29],"from":[30],"the":[31,112],"proliferation":[32],"of":[33,58],"visual":[34,81,98],"tokens":[35,82],"required":[36],"for":[37,96,157],"fine-grained":[38],"image":[39],"understanding.":[40],"We":[41],"propose":[42],"Skip-Vision,":[43],"a":[44,105,154],"unified":[45],"framework":[46],"addressing":[47],"both":[48],"inference":[51,101,134],"inefficiencies":[52],"vision-language":[54],"models.":[55],"On":[56],"top":[57],"conventional":[59],"token":[60],"compression":[61],"approaches,":[62],"our":[63,89],"method":[64],"introduces":[65],"two":[66],"complementary":[67],"acceleration":[68],"strategies.":[69],"For":[70,100],"acceleration,":[72,102],"we":[73,103],"observe":[74],"that":[75,110,125],"Feed-Forward":[76],"Network":[77],"(FFN)":[78],"computations":[79],"on":[80],"induce":[83],"marginal":[84],"feature":[85],"updates.":[86],"This":[87],"motivates":[88],"Skip-FFN":[90],"strategy,":[91],"which":[92],"bypasses":[93],"FFN":[94],"layers":[95],"redundant":[97],"tokens.":[99],"design":[104],"selective":[106],"KV-cache":[107],"removal":[108],"mechanism":[109],"prunes":[111],"skipped":[113],"key-value":[114],"pairs":[115],"during":[116],"decoding":[117],"while":[118,142],"preserving":[119],"performance.":[121],"Experimental":[122],"results":[123],"demonstrate":[124],"Skip-Vision":[126],"reduces":[127],"time":[129],"by":[130,136,140],"up":[131],"to":[132,148],"35\\%,":[133],"FLOPs":[135],"75\\%,":[137],"latency":[139],"45\\%,":[141],"achieving":[143],"comparable":[144],"or":[145],"superior":[146],"performance":[147],"existing":[149],"methods.":[150],"Our":[151],"work":[152],"provides":[153],"practical":[155],"solution":[156],"high-performance":[159],"MLLMs":[160],"with":[161],"enhanced":[162],"efficiency.":[163]},"counts_by_year":[],"updated_date":"2026-07-15T18:14:33.161393","created_date":"2025-10-10T00:00:00"}