{"id":"https://openalex.org/W7152605290","doi":"https://doi.org/10.48550/arxiv.2604.06515","title":"Efficient Quantization of Mixture-of-Experts with Theoretical Generalization Guarantees","display_name":"Efficient Quantization of Mixture-of-Experts with Theoretical Generalization Guarantees","publication_year":2026,"publication_date":"2026-04-07","ids":{"openalex":"https://openalex.org/W7152605290","doi":"https://doi.org/10.48550/arxiv.2604.06515"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.06515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06515","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,"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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.06515","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5034149210","display_name":"Mohammed Nowaz Rabbani Chowdhury","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chowdhury, Mohammed Nowaz Rabbani","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034513925","display_name":"Kaoutar El Maghraoui","orcid":"https://orcid.org/0000-0002-1967-8749"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Maghraoui, Kaoutar El","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109753352","display_name":"H. Tsai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tsai, Hsinyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082043392","display_name":"Naigang Wang","orcid":"https://orcid.org/0000-0001-7664-0061"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Naigang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014981734","display_name":"Geoffrey W. Burr","orcid":"https://orcid.org/0000-0001-5717-2549"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Burr, Geoffrey W.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133245521","display_name":"Liu Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5035442444","display_name":"M. Wang","orcid":"https://orcid.org/0000-0003-2408-8089"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Meng","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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.39649999141693115,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.39649999141693115,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.25380000472068787,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.047600001096725464,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.8119000196456909},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5810999870300293},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5318999886512756},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.5230000019073486},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.41839998960494995},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.32330000400543213}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8119000196456909},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6272000074386597},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5810999870300293},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5318999886512756},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5230000019073486},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5216000080108643},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.41839998960494995},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.32330000400543213},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.30559998750686646},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.29030001163482666},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.27649998664855957},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2743000090122223},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.27379998564720154},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.26809999346733093},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2547000050544739}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.06515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06515","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,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.06515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06515","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,"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":"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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Sparse":[0],"Mixture-of-Experts":[1],"(MoE)":[2],"allows":[3],"scaling":[4],"of":[5,17,28,77,83,134],"language":[6],"and":[7,72,125,173,190],"vision":[8],"models":[9],"efficiently":[10],"by":[11],"activating":[12],"only":[13,192],"a":[14,88],"small":[15],"subset":[16],"experts":[18,145,154],"per":[19],"input.":[20],"While":[21],"this":[22,44],"reduces":[23],"computation,":[24],"the":[25,74,81,132],"large":[26,156],"number":[27],"parameters":[29],"still":[30],"incurs":[31],"substantial":[32,67],"memory":[33],"overhead":[34,194],"during":[35,110],"inference.":[36],"Post-training":[37],"quantization":[38,48,82,133,152],"has":[39],"been":[40,60],"explored":[41],"to":[42,80,98,118,131,142,146],"address":[43],"issue.":[45],"Because":[46],"uniform":[47],"suffers":[49],"from":[50],"significant":[51],"accuracy":[52,181],"loss":[53],"at":[54],"low":[55],"bit-widths,":[56],"mixed-precision":[57],"methods":[58],"have":[59],"recently":[61],"explored;":[62],"however,":[63],"they":[64],"often":[65],"require":[66],"computation":[68],"for":[69,195],"bit-width":[70,97,196],"allocation":[71],"overlook":[73],"varying":[75],"sensitivity":[76],"model":[78,126],"performance":[79,127],"different":[84],"experts.":[85],"We":[86],"propose":[87],"theoretically":[89],"grounded":[90],"expert-wise":[91],"mixed":[92],"precision":[93,148],"strategy":[94],"that":[95,149,176],"assigns":[96],"each":[99],"expert":[100],"primarily":[101],"based":[102],"on":[103,166],"their":[104],"change":[105],"in":[106],"routers":[107],"l2":[108],"norm":[109],"training.":[111],"Experts":[112],"with":[113,155],"smaller":[114],"changes":[115],"are":[116,160],"shown":[117],"capture":[119],"less":[120],"frequent":[121],"but":[122],"critical":[123],"features,":[124],"is":[128],"more":[129],"sensitive":[130],"these":[135],"experts,":[136],"thus":[137],"requiring":[138],"higher":[139,163,180],"precision.":[140,164],"Furthermore,":[141],"avoid":[143],"allocating":[144],"lower":[147],"inject":[150],"high":[151],"noise,":[153],"maximum":[157],"intra-neuron":[158],"variance":[159],"also":[161,186],"allocated":[162],"Experiments":[165],"large-scale":[167],"MoE":[168],"models,":[169],"including":[170],"Switch":[171],"Transformer":[172],"Mixtral,":[174],"show":[175],"our":[177],"method":[178],"achieves":[179],"than":[182],"existing":[183],"approaches,":[184],"while":[185],"reducing":[187],"inference":[188],"cost":[189],"incurring":[191],"negligible":[193],"assignment.":[197]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-10T00:00:00"}