{"id":"https://openalex.org/W7204943389","doi":"https://doi.org/10.48550/arxiv.2608.28726","title":"Pro-Router: Token-Aware Progressive Model Routing with Adaptive Edge-Cloud Collaboration for Efficient Multimodal LLM Inference","display_name":"Pro-Router: Token-Aware Progressive Model Routing with Adaptive Edge-Cloud Collaboration for Efficient Multimodal LLM Inference","publication_year":2026,"publication_date":"2026-08-28","ids":{"openalex":"https://openalex.org/W7204943389","doi":"https://doi.org/10.48550/arxiv.2608.28726"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.28726","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.28726","pdf_url":"https://arxiv.org/pdf/2608.28726","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.28726","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148896900","display_name":"Xinyuan Gui","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gui, Xinyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5143519836","display_name":"Shaowen Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99041443","display_name":"Mississippi State University","ror":"https://ror.org/0432jq872","country_code":"US","type":"education","lineage":["https://openalex.org/I4210141039","https://openalex.org/I99041443"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wang, Shaowen","raw_affiliation_strings":["Department of Computer Science and Engineering , Mississippi State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , Mississippi State University","institution_ids":["https://openalex.org/I99041443"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148922030","display_name":"Sheng Sun","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Sun, Sheng","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140184588","display_name":"Zijian Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zijian","raw_affiliation_strings":["Institute of AI for Industries , Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of AI for Industries , Chinese Academy of Sciences","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019165823","display_name":"Zishu Yu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yu, Zishu","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5122530860","display_name":"Zheming Yang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang, Zheming","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.6518999934196472,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.6518999934196472,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.13979999721050262,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.026900000870227814,"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/model-routing","display_name":"model routing","score":0.828000009059906},{"id":"https://openalex.org/keywords/multimodal-large-language-models","display_name":"multimodal large language models","score":0.9269999861717224},{"id":"https://openalex.org/keywords/cloud-edge-collaboration","display_name":"cloud-edge collaboration","score":0.9089999794960022},{"id":"https://openalex.org/keywords/inference-efficiency","display_name":"inference efficiency","score":0.5199999809265137}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7731999754905701},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6869000196456909},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.6507999897003174},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.5947999954223633},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.5863000154495239},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4487000107765198},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.4408000111579895},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.43320000171661377},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.4104999899864197},{"id":"https://openalex.org/C204948658","wikidata":"https://www.wikidata.org/wiki/Q1119410","display_name":"Static routing","level":4,"score":0.3714999854564667},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3700999915599823},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3402999937534332},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.334199994802475},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.32170000672340393},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.321399986743927},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.3149999976158142},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.31189998984336853},{"id":"https://openalex.org/C104954878","wikidata":"https://www.wikidata.org/wiki/Q1648707","display_name":"Routing protocol","level":3,"score":0.3084000051021576},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.29910001158714294},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2946999967098236},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28940001130104065},{"id":"https://openalex.org/C184896649","wikidata":"https://www.wikidata.org/wiki/Q290066","display_name":"Routing table","level":4,"score":0.28780001401901245},{"id":"https://openalex.org/C9659607","wikidata":"https://www.wikidata.org/wiki/Q1268903","display_name":"Dynamic Source Routing","level":4,"score":0.2838999927043915},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2784000039100647},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.2757999897003174},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.26820001006126404},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.25040000677108765},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.28726","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.28726","pdf_url":"https://arxiv.org/pdf/2608.28726","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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.2608.28726","doi":"https://doi.org/10.48550/arxiv.2608.28726","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.28726","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.28726","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.28726","pdf_url":"https://arxiv.org/pdf/2608.28726","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204943389.pdf","grobid_xml":"https://content.openalex.org/works/W7204943389.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7221126440","https://openalex.org/W7161006313","https://openalex.org/W7125802482","https://openalex.org/W7141371174","https://openalex.org/W4412171515","https://openalex.org/W4416247041","https://openalex.org/W7105237345","https://openalex.org/W7134073568","https://openalex.org/W4416237770","https://openalex.org/W4404783163"],"abstract_inverted_index":{"The":[0],"remarkable":[1],"performance":[2],"of":[3,13,119,202],"multimodal":[4,80,194],"large":[5],"language":[6,43],"models":[7,198],"(MLLMs)":[8],"comes":[9],"at":[10,243],"the":[11,48,52,115,122,127,139,145,169,172,187,200,210,234],"cost":[12,24],"substantial":[14],"computational":[15],"overhead,":[16],"posing":[17],"significant":[18],"challenges":[19],"to":[20,46,107,134,144,161,205],"real-time":[21],"deployment":[22],"and":[23,171,182,197,214],"effectiveness.":[25],"Existing":[26],"model":[27,44,71,124,236],"routing":[28,72,212,216,237],"approaches":[29],"either":[30],"decide":[31],"from":[32],"coarse":[33],"request-level":[34],"features":[35],"alone":[36],"or":[37,40,142],"spend":[38],"one":[39],"several":[41],"extra":[42],"passes":[45],"inspect":[47],"generated":[49],"response,":[50],"leaving":[51],"token-level":[53],"uncertainty":[54],"signals":[55],"that":[56,157],"emerge":[57],"during":[58],"generation":[59,101],"unused.":[60],"To":[61],"address":[62],"these":[63],"limitations,":[64],"we":[65,150],"propose":[66],"Pro-Router,":[67],"a":[68,85,91,111],"token-aware":[69,112],"progressive":[70,87],"method":[73],"with":[74],"adaptive":[75,153],"edge-cloud":[76,154],"collaboration":[77],"for":[78],"efficient":[79],"LLM":[81],"inference.":[82],"Pro-Router":[83],"employs":[84],"two-stage":[86],"decision":[88],"mechanism.":[89],"First,":[90],"lightweight":[92],"prompt":[93],"pre-scorer":[94],"module":[95],"performs":[96],"rapid":[97],"pre-screening":[98],"before":[99],"token":[100,121],"begins,":[102],"guiding":[103],"apparently":[104],"simple":[105],"requests":[106],"small":[108,123],"models.":[109],"Second,":[110],"verifier":[113],"reads":[114],"sampling":[116],"probability":[117],"distribution":[118],"each":[120,162],"generates,":[125],"estimating":[126],"model's":[128],"confidence":[129],"in":[130],"its":[131],"own":[132],"output":[133],"determine,":[135],"per":[136],"request,":[137],"whether":[138],"answer":[140],"ships":[141],"escalates":[143],"cloud-based":[146],"high-precision":[147],"model.":[148],"Furthermore,":[149],"design":[151],"an":[152],"serving":[155,223],"pipeline":[156,224],"sizes":[158],"every":[159],"dispatch":[160],"device's":[163],"measured":[164],"service":[165],"rate,":[166],"so":[167],"both":[168],"edge":[170],"cloud":[173],"tiers":[174],"stay":[175],"fully":[176],"utilized":[177],"without":[178],"manual":[179],"parameter":[180],"tuning":[181],"are":[183],"not":[184],"impacted":[185],"by":[186,218],"network":[188],"latency.":[189],"Extensive":[190],"experiments":[191],"on":[192],"multiple":[193],"benchmark":[195],"datasets":[196],"demonstrate":[199],"effectiveness":[201],"Pro-Router.":[203],"Compared":[204],"other":[206],"methods,":[207],"it":[208],"achieves":[209],"highest":[211],"accuracy":[213],"improves":[215],"speed":[217],"more":[219,227],"than":[220,228,233],"10x.":[221],"Its":[222],"also":[225],"reaches":[226],"75%":[229],"higher":[230],"end-to-end":[231],"throughput":[232],"existing":[235],"pipeline.":[238],"Our":[239],"code":[240],"is":[241],"available":[242],"https://github.com/xinyuangui2/pro-router.":[244]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-02T00:00:00"}