{"id":"https://openalex.org/W7160273421","doi":"https://doi.org/10.48550/arxiv.2605.02218","title":"CoVSpec: Efficient Device-Edge Co-Inference for Vision-Language Models via Speculative Decoding","display_name":"CoVSpec: Efficient Device-Edge Co-Inference for Vision-Language Models via Speculative Decoding","publication_year":2026,"publication_date":"2026-05-04","ids":{"openalex":"https://openalex.org/W7160273421","doi":"https://doi.org/10.48550/arxiv.2605.02218"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.02218","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02218","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.02218","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135391381","display_name":"Yuanyuan Jia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jia, Yuanyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054443310","display_name":"Shunpu Tang","orcid":"https://orcid.org/0000-0002-3126-7731"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Shunpu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135291359","display_name":"Qianqian Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Qianqian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8464999794960022,"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":0.8464999794960022,"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":0.06129999831318855,"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":0.018400000408291817,"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/decoding-methods","display_name":"Decoding methods","score":0.637499988079071},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5960000157356262},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.499099999666214},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.4690000116825104},{"id":"https://openalex.org/keywords/mobile-device","display_name":"Mobile device","score":0.4462999999523163},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4083000123500824},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4074999988079071},{"id":"https://openalex.org/keywords/enhanced-data-rates-for-gsm-evolution","display_name":"Enhanced Data Rates for GSM Evolution","score":0.4027000069618225},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.3736000061035156},{"id":"https://openalex.org/keywords/inefficiency","display_name":"Inefficiency","score":0.35190001130104065}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8618999719619751},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.637499988079071},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5960000157356262},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.499099999666214},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.4690000116825104},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.4462999999523163},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.426800012588501},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4083000123500824},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4074999988079071},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.4027000069618225},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3736000061035156},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.352400004863739},{"id":"https://openalex.org/C2778869765","wikidata":"https://www.wikidata.org/wiki/Q6028363","display_name":"Inefficiency","level":2,"score":0.35190001130104065},{"id":"https://openalex.org/C2780186347","wikidata":"https://www.wikidata.org/wiki/Q11414","display_name":"Subnetwork","level":2,"score":0.33820000290870667},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.3206999897956848},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3199999928474426},{"id":"https://openalex.org/C761482","wikidata":"https://www.wikidata.org/wiki/Q118093","display_name":"Transmission (telecommunications)","level":2,"score":0.3003000020980835},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.3003000020980835},{"id":"https://openalex.org/C193969084","wikidata":"https://www.wikidata.org/wiki/Q7452500","display_name":"Sequential decoding","level":4,"score":0.2957000136375427},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.2888999879360199},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.2809999883174896},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2809999883174896},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.28060001134872437},{"id":"https://openalex.org/C95491727","wikidata":"https://www.wikidata.org/wiki/Q992968","display_name":"Mobile telephony","level":3,"score":0.2777000069618225},{"id":"https://openalex.org/C94361409","wikidata":"https://www.wikidata.org/wiki/Q7882500","display_name":"Uncertainty reduction theory","level":2,"score":0.26460000872612},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.26249998807907104},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2621000111103058},{"id":"https://openalex.org/C2779971761","wikidata":"https://www.wikidata.org/wiki/Q629872","display_name":"Arbiter","level":2,"score":0.2612000107765198},{"id":"https://openalex.org/C158156997","wikidata":"https://www.wikidata.org/wiki/Q1416645","display_name":"Models of communication","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C24858836","wikidata":"https://www.wikidata.org/wiki/Q844718","display_name":"Theory of computation","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.2540000081062317},{"id":"https://openalex.org/C2778456923","wikidata":"https://www.wikidata.org/wiki/Q5337692","display_name":"Edge computing","level":3,"score":0.25290000438690186}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.02218","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02218","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.02218","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02218","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"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":{"Vision-language":[0],"models":[1],"(VLMs)":[2],"have":[3],"demonstrated":[4],"strong":[5],"capabilities":[6],"in":[7],"multimodal":[8],"perception":[9],"and":[10,26,73,119,136,158,179],"reasoning.":[11],"However,":[12],"deploying":[13],"large":[14],"VLMs":[15,63],"on":[16,40,50,108,164],"mobile":[17,42,110],"devices":[18],"remains":[19],"challenging":[20],"due":[21,68],"to":[22,62,69,151,172],"their":[23],"substantial":[24],"computational":[25],"memory":[27],"demands.":[28],"A":[29],"practical":[30],"alternative":[31],"is":[32],"device-edge":[33],"co-inference,":[34],"where":[35],"a":[36,46,97,144],"lightweight":[37],"draft":[38,138],"VLM":[39,49,91],"the":[41,51,109,133,137],"device":[43,111],"collaborates":[44],"with":[45,148,188],"larger":[47],"target":[48],"edge":[52],"server":[53],"via":[54],"speculative":[55,60,87],"decoding.":[56],"Nevertheless,":[57],"directly":[58],"extending":[59],"decoding":[61,88],"suffers":[64],"from":[65],"severe":[66],"inefficiency":[67],"excessive":[70],"visual-token":[71],"computation":[72],"high":[74],"communication":[75,181],"overhead.":[76,162],"To":[77],"address":[78],"these":[79],"challenges,":[80],"we":[81,94,123,142],"propose":[82],"CoVSpec,":[83],"an":[84,125],"efficient":[85],"collaborative":[86],"framework":[89,102],"for":[90],"inference.":[92],"Specifically,":[93],"first":[95],"develop":[96],"training-free":[98],"visual":[99,106],"token":[100,117],"reduction":[101],"that":[103,129,168],"prunes":[104],"redundant":[105],"tokens":[107],"by":[112,183],"jointly":[113],"considering":[114],"query":[115],"relevance,":[116],"activity,":[118],"low-rank":[120],"dependency.":[121],"Moreover,":[122],"design":[124],"adaptive":[126],"drafting":[127],"strategy":[128],"dynamically":[130],"adjusts":[131],"both":[132],"verification":[134,157],"frequency":[135],"length.":[139],"In":[140],"addition,":[141],"introduce":[143],"parallel":[145],"branching":[146],"mechanism":[147],"decoupled":[149],"verification-correction":[150],"improve":[152],"draft-side":[153],"utilization":[154],"during":[155],"target-side":[156],"reduce":[159],"correction-related":[160],"transmission":[161],"Experiments":[163],"multiple":[165],"benchmarks":[166],"show":[167],"CoVSpec":[169],"achieves":[170],"up":[171],"2.21x":[173],"higher":[174],"throughput":[175],"than":[176,185],"target-only":[177],"inference":[178],"reduces":[180],"overhead":[182],"more":[184],"96%":[186],"compared":[187],"baselines,":[189],"without":[190],"compromising":[191],"task":[192],"accuracy.":[193]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-06T00:00:00"}