{"id":"https://openalex.org/W4407422870","doi":"https://doi.org/10.48550/arxiv.2502.06832","title":"Optimizing Robustness and Accuracy in Mixture of Experts: A Dual-Model Approach","display_name":"Optimizing Robustness and Accuracy in Mixture of Experts: A Dual-Model Approach","publication_year":2025,"publication_date":"2025-02-05","ids":{"openalex":"https://openalex.org/W4407422870","doi":"https://doi.org/10.48550/arxiv.2502.06832"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2502.06832","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.06832","pdf_url":"https://arxiv.org/pdf/2502.06832","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2502.06832","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5016089743","display_name":"Xianghua Zhang","orcid":"https://orcid.org/0000-0002-2180-6543"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhang, Xu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102775611","display_name":"Kaidi Xu","orcid":"https://orcid.org/0000-0003-4437-0671"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Kaidi","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104223650","display_name":"Ziqing Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Ziqing","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100525902","display_name":"Ren Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ren","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5016089743"],"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/T11918","display_name":"Forecasting Techniques and Applications","score":0.7570000290870667,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T11918","display_name":"Forecasting Techniques and Applications","score":0.7570000290870667,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11502","display_name":"Facility Location and Emergency Management","score":0.7335000038146973,"subfield":{"id":"https://openalex.org/subfields/1407","display_name":"Organizational Behavior and Human Resource Management"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T13274","display_name":"Expert finding and Q&A systems","score":0.696399986743927,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.7548000812530518},{"id":"https://openalex.org/keywords/dual","display_name":"Dual (grammatical number)","score":0.6562240123748779},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5629265308380127},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3686733543872833},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.3371272385120392},{"id":"https://openalex.org/keywords/econometrics","display_name":"Econometrics","score":0.3278544843196869},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.22311261296272278},{"id":"https://openalex.org/keywords/art","display_name":"Art","score":0.058658748865127563},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.05854341387748718}],"concepts":[{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.7548000812530518},{"id":"https://openalex.org/C2780980858","wikidata":"https://www.wikidata.org/wiki/Q110022","display_name":"Dual (grammatical number)","level":2,"score":0.6562240123748779},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5629265308380127},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3686733543872833},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3371272385120392},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.3278544843196869},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.22311261296272278},{"id":"https://openalex.org/C142362112","wikidata":"https://www.wikidata.org/wiki/Q735","display_name":"Art","level":0,"score":0.058658748865127563},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.05854341387748718},{"id":"https://openalex.org/C55493867","wikidata":"https://www.wikidata.org/wiki/Q7094","display_name":"Biochemistry","level":1,"score":0.0},{"id":"https://openalex.org/C124952713","wikidata":"https://www.wikidata.org/wiki/Q8242","display_name":"Literature","level":1,"score":0.0},{"id":"https://openalex.org/C104317684","wikidata":"https://www.wikidata.org/wiki/Q7187","display_name":"Gene","level":2,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2502.06832","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.06832","pdf_url":"https://arxiv.org/pdf/2502.06832","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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.2502.06832","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2502.06832","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":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2502.06832","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.06832","pdf_url":"https://arxiv.org/pdf/2502.06832","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":[],"awards":[{"id":"https://openalex.org/G2888593251","display_name":null,"funder_award_id":"2246157","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G542394987","display_name":null,"funder_award_id":"2319243","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320338284","display_name":"Argonne National Laboratory","ror":"https://ror.org/05gvnxz63"}],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4407422870.pdf"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2961085424","https://openalex.org/W4306674287","https://openalex.org/W4387369504","https://openalex.org/W3046775127","https://openalex.org/W4394896187","https://openalex.org/W3170094116","https://openalex.org/W4386462264","https://openalex.org/W3107602296","https://openalex.org/W4364306694","https://openalex.org/W4312192474"],"abstract_inverted_index":{"Mixture":[0],"of":[1,39,57,96,102,168,215],"Experts":[2],"(MoE)":[3],"have":[4],"shown":[5],"remarkable":[6],"success":[7],"in":[8,30],"leveraging":[9],"specialized":[10],"expert":[11,62,105],"networks":[12,63],"for":[13,28,139,156,180],"complex":[14],"machine":[15],"learning":[16],"tasks.":[17],"However,":[18],"their":[19],"susceptibility":[20],"to":[21,41,68,91],"adversarial":[22,69,94],"attacks":[23,70],"presents":[24],"a":[25,80,87,117,123,133,174],"critical":[26,37],"challenge":[27],"deployment":[29],"robust":[31,82],"applications.":[32],"This":[33,136,183],"paper":[34],"addresses":[35],"the":[36,55,72,93,100,143,158,162,166,181,213],"question":[38],"how":[40],"incorporate":[42],"robustness":[43,95,154,169,188],"into":[44],"MoEs":[45],"while":[46],"maintaining":[47],"high":[48],"natural":[49],"accuracy.":[50],"We":[51,146],"begin":[52],"by":[53,151],"analyzing":[54],"vulnerability":[56],"MoE":[58,125,130,160],"components,":[59],"finding":[60],"that":[61,85,120],"are":[64],"notably":[65],"more":[66],"susceptible":[67],"than":[71],"router.":[73],"Based":[74],"on":[75,113,200],"this":[76],"insight,":[77],"we":[78,115,172],"propose":[79,173],"targeted":[81],"training":[83,108,177,185],"technique":[84],"integrates":[86],"novel":[88,175],"loss":[89],"function":[90],"enhance":[92],"MoE,":[97],"requiring":[98],"only":[99],"robustification":[101],"one":[103],"additional":[104],"without":[106],"compromising":[107],"or":[109],"inference":[110],"efficiency.":[111],"Building":[112],"this,":[114],"introduce":[116],"dual-model":[118],"strategy":[119,178],"linearly":[121],"combines":[122],"standard":[124],"model":[126,131],"with":[127,195],"our":[128,216],"robustified":[129],"using":[132,205],"smoothing":[134],"parameter.":[135],"approach":[137],"allows":[138],"flexible":[140],"control":[141],"over":[142],"robustness-accuracy":[144],"trade-off.":[145],"further":[147],"provide":[148],"theoretical":[149],"foundations":[150],"deriving":[152],"certified":[153],"bounds":[155],"both":[157,187],"single":[159],"and":[161,170,189,202,207],"dual-model.":[163,182],"To":[164],"push":[165],"boundaries":[167],"accuracy,":[171],"joint":[176,184],"JTDMoE":[179],"enhances":[186],"accuracy":[190],"beyond":[191],"what":[192],"is":[193,221],"achievable":[194],"separate":[196],"models.":[197],"Experimental":[198],"results":[199],"CIFAR-10":[201],"TinyImageNet":[203],"datasets":[204],"ResNet18":[206],"Vision":[208],"Transformer":[209],"(ViT)":[210],"architectures":[211],"demonstrate":[212],"effectiveness":[214],"proposed":[217],"methods.":[218],"The":[219],"code":[220],"publicly":[222],"available":[223],"at":[224],"https://github.com/TIML-Group/Robust-MoE-Dual-Model.":[225]},"counts_by_year":[],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-10T00:00:00"}