{"id":"https://openalex.org/W7138268053","doi":"https://doi.org/10.1609/aaai.v40i31.39816","title":"CasMoE: A Cascaded Framework for Efficient MoE Inference on Resource-constrained Devices","display_name":"CasMoE: A Cascaded Framework for Efficient MoE Inference on Resource-constrained Devices","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138268053","doi":"https://doi.org/10.1609/aaai.v40i31.39816"},"language":"en","primary_location":{"id":"doi:10.1609/aaai.v40i31.39816","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i31.39816","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i31.39816","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100307786","display_name":"Chengcheng Wang","orcid":"https://orcid.org/0009-0000-6974-1358"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chengcheng Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129706789","display_name":"Haowen He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haowen He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129656769","display_name":"Liang Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129669775","display_name":"Xiaoheng Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaoheng Deng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129683431","display_name":"Lixin Duan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lixin Duan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5129661619","display_name":"Shaohua Wan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shaohua Wan","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":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"31","first_page":"26133","last_page":"26141"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.27140000462532043,"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"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.27140000462532043,"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"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.10729999840259552,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.07840000092983246,"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":0.7689999938011169},{"id":"https://openalex.org/keywords/expert-system","display_name":"Expert system","score":0.5846999883651733},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5794000029563904},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5497000217437744},{"id":"https://openalex.org/keywords/parameterized-complexity","display_name":"Parameterized complexity","score":0.4884999990463257},{"id":"https://openalex.org/keywords/inference-engine","display_name":"Inference engine","score":0.4702000021934509},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.38499999046325684}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7689999938011169},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7555000185966492},{"id":"https://openalex.org/C58328972","wikidata":"https://www.wikidata.org/wiki/Q184609","display_name":"Expert system","level":2,"score":0.5846999883651733},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5794000029563904},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5497000217437744},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5105000138282776},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4934000074863434},{"id":"https://openalex.org/C165464430","wikidata":"https://www.wikidata.org/wiki/Q1570441","display_name":"Parameterized complexity","level":2,"score":0.4884999990463257},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.4702000021934509},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.38499999046325684},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.37689998745918274},{"id":"https://openalex.org/C105002631","wikidata":"https://www.wikidata.org/wiki/Q4833645","display_name":"Subject-matter expert","level":3,"score":0.3596999943256378},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.3589000105857849},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.3393999934196472},{"id":"https://openalex.org/C2777851325","wikidata":"https://www.wikidata.org/wiki/Q7094102","display_name":"Online model","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.30070000886917114},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2987000048160553},{"id":"https://openalex.org/C86610423","wikidata":"https://www.wikidata.org/wiki/Q1925081","display_name":"Metamodeling","level":2,"score":0.25780001282691956},{"id":"https://openalex.org/C155846161","wikidata":"https://www.wikidata.org/wiki/Q1143367","display_name":"Graphical model","level":2,"score":0.25200000405311584}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1609/aaai.v40i31.39816","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i31.39816","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},{"id":"pmh:oai:ojs.aaai.org:article/39816","is_oa":false,"landing_page_url":"https://ojs.aaai.org/index.php/AAAI/article/view/39816","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2159-5399","raw_type":"info:eu-repo/semantics/article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i31.39816","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i31.39816","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3494558276","display_name":null,"funder_award_id":"62172438","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5022004486","display_name":null,"funder_award_id":"2024A1515011155","funder_id":"https://openalex.org/F4320321921","funder_display_name":"Natural Science Foundation of Guangdong Province"},{"id":"https://openalex.org/G815783936","display_name":null,"funder_award_id":"62172438","funder_id":"https://openalex.org/F4320321921","funder_display_name":"Natural Science Foundation of Guangdong Province"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321921","display_name":"Natural Science Foundation of Guangdong Province","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"Mixture-of-Experts":[1],"(MoE)":[2],"architecture":[3],"has":[4],"emerged":[5],"as":[6,50],"a":[7,74,89,102,125,143,161,184],"key":[8],"enabler":[9],"for":[10,80,156],"scaling":[11],"large":[12],"language":[13],"models":[14],"(LLMs),":[15],"empowering":[16],"increased":[17,66],"model":[18,51],"capacity":[19],"with":[20,138],"minimal":[21],"computational":[22],"overhead":[23],"through":[24],"gating-based":[25],"dynamic":[26],"expert":[27,37,96,115,133],"activation.":[28],"However,":[29],"due":[30],"to":[31,62,93,110,141,165,173],"the":[32,99,113,118,122,139,150,176,179],"memory":[33],"demands":[34],"introduced":[35,109],"by":[36],"modules,":[38],"MoE":[39,82,151,158],"inference":[40,83,188,203],"on":[41,84,194],"resource-constrained":[42,85],"devices":[43],"is":[44,108,135,170],"still":[45],"challenging.":[46],"Existing":[47],"methods":[48],"such":[49],"compression":[52],"and":[53,76,181,190],"parameter":[54],"offloading":[55],"provide":[56],"partial":[57],"alleviation":[58],"but":[59],"often":[60],"lead":[61],"reduced":[63],"accuracy":[64],"or":[65],"latency.":[67],"In":[68,98,121],"this":[69],"paper,":[70],"we":[71],"propose":[72],"CasMoE,":[73],"general":[75],"efficient":[77,95],"cascaded":[78],"framework":[79],"accelerating":[81,202],"devices.":[86],"CasMoE":[87],"employs":[88],"two-stage":[90],"offline-online":[91],"approach":[92],"facilitate":[94],"prefetching.":[97],"offline":[100],"stage,":[101,124],"parameterized":[103],"Expert":[104,127],"Activation":[105,128],"Predictor":[106],"(EAP)":[107],"accurately":[111],"predict":[112],"corresponding":[114],"activation":[116],"from":[117],"incoming":[119],"prompt.":[120],"online":[123],"non-parametric":[126],"Matcher":[129],"(EAM)":[130],"supporting":[131],"fast":[132],"retrieval":[134],"then":[136],"integrated":[137],"EAP":[140],"form":[142],"cascade":[144],"planner":[145],"that":[146],"operates":[147],"independently":[148],"of":[149,178],"architecture,":[152],"predicting":[153],"activated":[154],"experts":[155],"all":[157],"layers":[159],"in":[160,201],"single":[162],"pass":[163],"prior":[164],"decoding.":[166],"A":[167],"gating":[168],"mechanism":[169],"also":[171],"incorporated":[172],"dynamically":[174],"adjust":[175],"sensitivity":[177],"EAM":[180],"EAP,":[182],"enabling":[183],"flexible":[185],"trade-off":[186],"between":[187],"efficiency":[189],"quality.":[191],"Extensive":[192],"experiments":[193],"diverse":[195],"downstream":[196],"tasks":[197],"demonstrate":[198],"CasMoE\u2019s":[199],"effectiveness":[200],"while":[204],"preserving":[205],"high":[206],"accuracy.":[207]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-03-18T00:00:00"}