{"id":"https://openalex.org/W7164891892","doi":"https://doi.org/10.48550/arxiv.2606.15453","title":"A Spatio-Temporal Expert Prefetching Framework for Efficient MoE-based LLM Inference","display_name":"A Spatio-Temporal Expert Prefetching Framework for Efficient MoE-based LLM Inference","publication_year":2026,"publication_date":"2026-06-13","ids":{"openalex":"https://openalex.org/W7164891892","doi":"https://doi.org/10.48550/arxiv.2606.15453"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.15453","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15453","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.2606.15453","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5138721659","display_name":"Yingnan Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Yingnan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134460199","display_name":"Razvan Bunescu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bunescu, Razvan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034189643","display_name":"Ahmed Louri","orcid":"https://orcid.org/0000-0003-4262-6688"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Louri, Ahmed","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014895124","display_name":"Avinash Karanth","orcid":"https://orcid.org/0000-0002-9472-4637"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Karanth, Avinash","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138691629","display_name":"Ke Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ke","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/T14347","display_name":"Big Data and Digital Economy","score":0.13220000267028809,"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/T14347","display_name":"Big Data and Digital Economy","score":0.13220000267028809,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.08839999884366989,"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/T10028","display_name":"Topic Modeling","score":0.08799999952316284,"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/inference","display_name":"Inference","score":0.7279999852180481},{"id":"https://openalex.org/keywords/expert-system","display_name":"Expert system","score":0.6748999953269958},{"id":"https://openalex.org/keywords/subject-matter-expert","display_name":"Subject-matter expert","score":0.5898000001907349},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5368000268936157},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.4993000030517578},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.4560000002384186},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.43290001153945923},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.42250001430511475}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7713000178337097},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7279999852180481},{"id":"https://openalex.org/C58328972","wikidata":"https://www.wikidata.org/wiki/Q184609","display_name":"Expert system","level":2,"score":0.6748999953269958},{"id":"https://openalex.org/C105002631","wikidata":"https://www.wikidata.org/wiki/Q4833645","display_name":"Subject-matter expert","level":3,"score":0.5898000001907349},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5368000268936157},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5098000168800354},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4993000030517578},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.4560000002384186},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43619999289512634},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.43290001153945923},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.42250001430511475},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.4120999872684479},{"id":"https://openalex.org/C102600418","wikidata":"https://www.wikidata.org/wiki/Q6517507","display_name":"Legal expert system","level":3,"score":0.3919999897480011},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.38679999113082886},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3767000138759613},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.36250001192092896},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.36169999837875366},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3440000116825104},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.3409000039100647},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3075999915599823},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3057999908924103},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.305400013923645},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.2759999930858612},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.26339998841285706},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.2628999948501587}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.15453","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15453","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.2606.15453","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15453","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":[{"score":0.9022720456123352,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Mixture-of-Experts":[0],"(MoE)":[1],"based":[2],"large":[3],"language":[4,134],"models":[5,54],"(LLMs),":[6],"such":[7],"as":[8,15],"Qwen":[9],"and":[10,42,72,111,131,136,157,232],"DeepSeek,":[11],"have":[12],"recently":[13],"emerged":[14],"an":[16],"effective":[17],"approach":[18],"to":[19,95,185],"improving":[20],"model":[21,237],"capacity":[22],"without":[23],"proportionally":[24],"increasing":[25],"computational":[26],"cost.":[27],"By":[28],"replacing":[29],"the":[30,57,64,86,201,220,224],"conventional":[31],"feed-forward":[32],"network":[33],"in":[34,127],"dense":[35],"LLMs":[36,130],"with":[37,189,205,223],"a":[38,45,100,106,120,173,194,206],"set":[39],"of":[40,47,60,109,123,183,219],"experts":[41,88,181],"activating":[43],"only":[44],"subset":[46],"them":[48],"for":[49],"each":[50,144],"input":[51],"token,":[52],"MoE":[53,155,229],"significantly":[55,227],"increase":[56],"total":[58],"number":[59],"parameters":[61],"while":[62,235],"keeping":[63],"per-token":[65],"computation":[66],"relatively":[67],"manageable.":[68],"However,":[69],"this":[70,115,168],"dynamic":[71,213],"irregular":[73],"expert":[74,80,102,124,147,163,175,187,214],"activation":[75],"pattern":[76],"also":[77],"introduces":[78],"substantial":[79],"loading":[81,103,188],"overhead":[82],"during":[83],"inference,":[84],"since":[85],"required":[87],"must":[89],"be":[90],"fetched":[91],"on":[92,167],"demand":[93],"according":[94],"token-dependent":[96],"routing":[97,203],"results.":[98],"As":[99],"result,":[101],"latency":[104],"becomes":[105],"major":[107],"source":[108],"performance":[110,231],"energy":[112,233],"inefficiency.":[113],"To":[114],"end,":[116],"we":[117,170],"first":[118],"perform":[119],"comprehensive":[121],"analysis":[122,140],"selection":[125],"behavior":[126,204],"various":[128],"MoE-based":[129],"applications,":[132],"including":[133],"understanding":[135],"code":[137],"generation.":[138],"Our":[139],"reveals":[141],"that,":[142],"within":[143],"application":[145],"domain,":[146],"requests":[148],"exhibit":[149],"strong":[150],"correlation":[151],"across":[152],"both":[153],"adjacent":[154],"layers":[156],"consecutive":[158],"decoding":[159],"tokens,":[160],"making":[161],"future":[162],"activations":[164],"predictable.":[165],"Based":[166],"insight,":[169],"propose":[171],"ST-MoE,":[172],"spatio-temporal":[174],"prefetching":[176],"framework":[177],"that":[178,199,210],"proactively":[179],"stages":[180],"ahead":[182],"use":[184],"overlap":[186],"ongoing":[190],"computation.":[191],"ST-MoE":[192],"combines":[193],"lightweight":[195],"runtime":[196],"prediction":[197,221],"mechanism":[198,222],"preserves":[200],"original":[202],"reconfigurable":[207],"hardware":[208,226],"design":[209],"efficiently":[211],"supports":[212],"prefetching.":[215],"The":[216],"combined":[217],"effect":[218],"supporting":[225],"improves":[228],"inference":[230,238],"efficiency":[234],"preserving":[236],"accuracy.":[239]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-17T00:00:00"}