{"id":"https://openalex.org/W7225409378","doi":"https://doi.org/10.5281/zenodo.23998378","title":"Predictive Model Switching for Generative AI: Anticipating Future Model Demand to Coordinate Proactive Loading, Residency, and Unloading","display_name":"Predictive Model Switching for Generative AI: Anticipating Future Model Demand to Coordinate Proactive Loading, Residency, and Unloading","publication_year":2026,"publication_date":"2026-11-06","ids":{"openalex":"https://openalex.org/W7225409378","doi":"https://doi.org/10.5281/zenodo.23998378"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.23998378","doi":"https://doi.org/10.5281/zenodo.23998378","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23998378","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.23998378","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100638045","display_name":"Jincheng Zhang","orcid":"https://orcid.org/0000-0001-7332-6704","observed_orcids":["https://orcid.org/0000-0001-7332-6704","https://orcid.org/0009-0005-1860-0009"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhang, Jincheng","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0005-1860-0009","affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5100638045"],"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.8223000168800354,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.8223000168800354,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.06120000034570694,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.033799998462200165,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.921999990940094},{"id":"https://openalex.org/keywords/gpu-memory-utilization","display_name":"GPU memory utilization","score":0.7139999866485596},{"id":"https://openalex.org/keywords/cold-start-latency","display_name":"cold start latency","score":0.8320000171661377},{"id":"https://openalex.org/keywords/inference-latency","display_name":"inference latency","score":0.46799999475479126},{"id":"https://openalex.org/keywords/energy-consumption","display_name":"energy consumption","score":0.4449999928474426}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7799000144004822},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.47999998927116394},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.42719998955726624},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.41749998927116394},{"id":"https://openalex.org/C114466953","wikidata":"https://www.wikidata.org/wiki/Q6034165","display_name":"Initialization","level":2,"score":0.40369999408721924},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.385699987411499},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3801000118255615},{"id":"https://openalex.org/C13606891","wikidata":"https://www.wikidata.org/wiki/Q2623243","display_name":"Conceptual model","level":2,"score":0.3763999938964844},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.35989999771118164},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34549999237060547},{"id":"https://openalex.org/C2779714256","wikidata":"https://www.wikidata.org/wiki/Q25305062","display_name":"Multiple Models","level":2,"score":0.3440000116825104},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.34389999508857727},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.33820000290870667},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.3046000003814697},{"id":"https://openalex.org/C193809577","wikidata":"https://www.wikidata.org/wiki/Q3409300","display_name":"Demand forecasting","level":2,"score":0.2750000059604645},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2721000015735626},{"id":"https://openalex.org/C45804977","wikidata":"https://www.wikidata.org/wiki/Q7239673","display_name":"Predictive modelling","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.2565999925136566},{"id":"https://openalex.org/C53833338","wikidata":"https://www.wikidata.org/wiki/Q1061424","display_name":"Context switch","level":2,"score":0.2524000108242035},{"id":"https://openalex.org/C2776834041","wikidata":"https://www.wikidata.org/wiki/Q25346349","display_name":"Execution model","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.23998378","doi":"https://doi.org/10.5281/zenodo.23998378","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23998378","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},{"id":"doi:10.5281/zenodo.23998379","doi":"https://doi.org/10.5281/zenodo.23998379","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23998379","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.5281/zenodo.23998378","doi":"https://doi.org/10.5281/zenodo.23998378","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23998378","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7225400169","https://openalex.org/W7225413377","https://openalex.org/W7225385268","https://openalex.org/W7225411818","https://openalex.org/W7225392849","https://openalex.org/W7225432460","https://openalex.org/W7225408802","https://openalex.org/W7225424338","https://openalex.org/W7225267314","https://openalex.org/W7225375373"],"abstract_inverted_index":{"The":[0,256,353],"increasing":[1],"deployment":[2],"of":[3,30,162,411,441],"generative":[4,38],"artificial":[5],"intelligence":[6],"(AI)":[7],"systems":[8,57],"has":[9],"resulted":[10],"in":[11,46,65],"inference":[12,274,424,481],"infrastructures":[13],"supporting":[14],"multiple":[15],"models":[16,34,39,64,179,247,322,443],"with":[17,306,422],"heterogeneous":[18],"capabilities,":[19],"computational":[20,397,433],"requirements,":[21],"memory":[22,48,81,206,294,389,418],"footprints,":[23],"and":[24,36,70,93,116,150,176,216,229,276,300,321,372,376,401,420,437,475],"request":[25,126,289],"distributions.":[26],"As":[27],"the":[28,301,328,409,439,442,462],"number":[29],"available":[31,251],"large":[32],"language":[33],"(LLMs)":[35],"specialized":[37],"increases,":[40],"maintaining":[41],"every":[42],"model":[43,76,95,103,145,154,164,210,218,227,231,236,291,336,361,367,384,393,467,472],"permanently":[44],"resident":[45],"GPU":[47,80,205,268,388],"becomes":[49,105],"increasingly":[50],"expensive":[51],"or":[52,62,187,340,483],"technically":[53],"infeasible.":[54],"Conventional":[55],"model-serving":[56],"commonly":[58],"activate,":[59],"load,":[60],"swap,":[61],"unload":[63],"response":[66],"to":[67,330,408,445,491],"current":[68,168],"requests":[69,159,254],"observed":[71],"resource":[72,477],"utilization.":[73],"Although":[74],"dynamic":[75],"management":[77,469],"can":[78,147,470],"improve":[79,476],"utilization,":[82,390],"reactive":[83,360],"switching":[84,244,272,283,391,455],"may":[85,112],"introduce":[86],"model-loading":[87,202],"delays,":[88],"cold-start":[89],"latency,":[90,275,386],"memory-transfer":[91],"overhead,":[92],"unnecessary":[94],"replacement":[96],"during":[97,107],"rapidly":[98],"changing":[99],"workloads.":[100,448],"Furthermore,":[101,309],"a":[102,138,190,259,310,335],"that":[104,141,286,453],"necessary":[106],"an":[108,239,281],"upcoming":[109],"demand":[110,146,174,351],"surge":[111],"require":[113],"substantial":[114],"initialization":[115],"loading":[117,270,375,385],"time":[118,203],"if":[119],"preparation":[120],"begins":[121],"only":[122,333],"after":[123],"its":[124],"first":[125],"arrives.":[127],"This":[128,326],"paper":[129],"introduces":[130,280],"Predictive":[131],"Model":[132],"Switching":[133],"for":[134,262],"Generative":[135],"AI":[136],"(PMS-GAI),":[137],"conceptual":[139],"framework":[140,257,329],"investigates":[142,461],"whether":[143,334],"future":[144,172,253,264],"be":[148,181,250,338,347],"predicted":[149,288],"translated":[151],"into":[152],"proactive":[153,209],"lifecycle":[155,468],"decisions":[156],"before":[157,349],"corresponding":[158],"materialize.":[160],"Instead":[161],"determining":[163],"residency":[165,212,312,364],"exclusively":[166],"from":[167,430],"requests,":[169],"PMS-GAI":[170,426],"estimates":[171],"model-specific":[173],"distributions":[175],"anticipates":[177],"which":[178,235,246,465],"should":[180,237,248,337,346],"loaded,":[182],"retained,":[183],"partially":[184],"prepared,":[185],"migrated,":[186],"unloaded":[188],"within":[189],"given":[191,407],"prediction":[192,298,382,412],"horizon.":[193],"Its":[194,486],"architecture":[195,313],"integrates":[196],"predictive":[197,230,243,370,428,454],"model-demand":[198,381],"forecasting,":[199],"workload-to-model":[200],"mapping,":[201],"estimation,":[204],"availability":[207,440],"prediction,":[208],"activation,":[211],"management,":[213,365],"cost-aware":[214],"unloading,":[215,417],"adaptive":[217],"switching.":[219],"A":[220],"central":[221],"distinction":[222],"is":[223,406,456],"established":[224],"between":[225,359],"request-level":[226],"routing":[228,233],"switching:":[232],"determines":[234,245],"process":[238],"incoming":[240],"request,":[241],"whereas":[242],"already":[249],"when":[252],"arrive.":[255],"develops":[258],"mathematical":[260],"formulation":[261],"selecting":[263],"model-residency":[265],"configurations":[266],"under":[267,464],"memory,":[269],"bandwidth,":[271],"cost,":[273,398],"service-level":[277,402],"constraints.":[278],"It":[279,449],"expected":[282,302],"utility":[284],"function":[285],"considers":[287],"volume,":[290],"activation":[292,473],"costs,":[293,297],"occupation,":[295],"unloading":[296],"uncertainty,":[299],"latency":[303],"penalties":[304],"associated":[305],"unavailable":[307],"models.":[308],"multi-stage":[311],"distinguishes":[314],"GPU-resident":[315],"models,":[316,318,320],"host-memory-resident":[317],"storage-resident":[319],"requiring":[323],"remote":[324],"retrieval.":[325],"enables":[327],"investigate":[331],"not":[332,451],"loaded":[339],"unloaded,":[341],"but":[342],"also":[343],"where":[344],"it":[345,460],"maintained":[348],"anticipated":[350,432],"materializes.":[352],"proposed":[354],"methodology":[355],"establishes":[356],"experimental":[357],"comparisons":[358],"switching,":[362,415],"frequency-based":[363],"threshold-based":[366],"loading,":[368],"conventional":[369],"preloading,":[371],"prediction-driven":[373],"joint":[374],"unloading.":[377],"Evaluation":[378],"metrics":[379],"include":[380],"accuracy,":[383],"time-to-first-token,":[387],"frequency,":[392],"eviction":[394],"rate,":[395],"total":[396],"energy":[399],"consumption,":[400],"compliance.":[403],"Particular":[404],"attention":[405],"risks":[410],"errors,":[413],"excessive":[414],"premature":[416],"contention,":[419],"interference":[421],"active":[423],"sessions.":[425],"extends":[427],"computing":[429],"preparing":[431],"workloads":[434],"toward":[435],"anticipating":[436],"managing":[438],"required":[444],"execute":[446],"those":[447],"does":[450],"assume":[452],"universally":[457],"beneficial;":[458],"rather,":[459],"conditions":[463],"future-aware":[466],"reduce":[471],"overhead":[474],"utilization":[478],"without":[479],"compromising":[480],"quality":[482],"service":[484],"reliability.":[485],"practical":[487],"effectiveness":[488],"remains":[489],"subject":[490],"empirical":[492],"validation.":[493]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-11-07T00:00:00"}