{"id":"https://openalex.org/W7214947677","doi":"https://doi.org/10.48550/arxiv.2609.36301","title":"MoRE: Scaling mixture of experts with hardware-aware low-rank routing","display_name":"MoRE: Scaling mixture of experts with hardware-aware low-rank routing","publication_year":2026,"publication_date":"2026-09-28","ids":{"openalex":"https://openalex.org/W7214947677","doi":"https://doi.org/10.48550/arxiv.2609.36301"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.36301","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36301","pdf_url":"https://arxiv.org/pdf/2609.36301","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.36301","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101251631","display_name":"Honam Wong","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wong, Honam","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019144684","display_name":"Surbhi Goel","orcid":"https://orcid.org/0000-0002-0385-9043","observed_orcids":["https://orcid.org/0000-0002-0385-9043"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Goel, Surbhi","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5091031925","display_name":"Enric Boix-Adser\u00e0","orcid":"https://orcid.org/0000-0003-0635-9703","observed_orcids":["https://orcid.org/0000-0003-0635-9703"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Boix-Adser\u00e0, Enric","raw_affiliation_strings":["The Wharton School , University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Wharton School , University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I79576946"],"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/T10036","display_name":"Advanced Neural Network Applications","score":0.4375,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.4375,"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":0.3693999946117401,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.05389999970793724,"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/mixture-of-experts","display_name":"mixture of experts","score":0.9909999966621399},{"id":"https://openalex.org/keywords/routing-cost-reduction","display_name":"routing cost reduction","score":0.6430000066757202},{"id":"https://openalex.org/keywords/high-bandwidth-memory","display_name":"high-bandwidth memory","score":0.7749999761581421},{"id":"https://openalex.org/keywords/memorization","display_name":"memorization","score":0.6880000233650208},{"id":"https://openalex.org/keywords/knowledge-intensive-question-answering","display_name":"knowledge-intensive question answering","score":0.6800000071525574},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.5809999704360962},{"id":"https://openalex.org/keywords/load-balancing","display_name":"load balancing","score":0.652999997138977}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6823999881744385},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.5491999983787537},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5321000218391418},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.4731999933719635},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.43459999561309814},{"id":"https://openalex.org/C39927690","wikidata":"https://www.wikidata.org/wiki/Q11197","display_name":"Logarithm","level":2,"score":0.420199990272522},{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.41339999437332153},{"id":"https://openalex.org/C2775896111","wikidata":"https://www.wikidata.org/wiki/Q642560","display_name":"Router","level":2,"score":0.4059000015258789},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3926999866962433},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.37940001487731934},{"id":"https://openalex.org/C33676613","wikidata":"https://www.wikidata.org/wiki/Q13415176","display_name":"Dimension (graph theory)","level":2,"score":0.37779998779296875},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.3723999857902527},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.36880001425743103},{"id":"https://openalex.org/C42747912","wikidata":"https://www.wikidata.org/wiki/Q1048447","display_name":"Multiplicative function","level":2,"score":0.3634999990463257},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.36320000886917114},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35120001435279846},{"id":"https://openalex.org/C159254197","wikidata":"https://www.wikidata.org/wiki/Q1144915","display_name":"Lexicographical order","level":2,"score":0.34279999136924744},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.34130001068115234},{"id":"https://openalex.org/C86037889","wikidata":"https://www.wikidata.org/wiki/Q4330127","display_name":"Learning to rank","level":3,"score":0.3402999937534332},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3303000032901764},{"id":"https://openalex.org/C184992742","wikidata":"https://www.wikidata.org/wiki/Q7243229","display_name":"Prime (order theory)","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.311599999666214},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C58166","wikidata":"https://www.wikidata.org/wiki/Q224821","display_name":"Fuzzy logic","level":2,"score":0.3093000054359436},{"id":"https://openalex.org/C187834632","wikidata":"https://www.wikidata.org/wiki/Q188804","display_name":"Factorization","level":2,"score":0.3075000047683716},{"id":"https://openalex.org/C2781039887","wikidata":"https://www.wikidata.org/wiki/Q1391724","display_name":"Factor (programming language)","level":2,"score":0.30230000615119934},{"id":"https://openalex.org/C184896649","wikidata":"https://www.wikidata.org/wiki/Q290066","display_name":"Routing table","level":4,"score":0.2921999990940094},{"id":"https://openalex.org/C199422724","wikidata":"https://www.wikidata.org/wiki/Q41118","display_name":"Quotient","level":2,"score":0.28760001063346863},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.2752000093460083},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.2752000093460083},{"id":"https://openalex.org/C43364308","wikidata":"https://www.wikidata.org/wiki/Q8799","display_name":"Byte","level":2,"score":0.2727000117301941},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.262800008058548},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.25589999556541443},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.2556000053882599},{"id":"https://openalex.org/C2780861071","wikidata":"https://www.wikidata.org/wiki/Q1062934","display_name":"Character (mathematics)","level":2,"score":0.251800000667572},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.36301","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36301","pdf_url":"https://arxiv.org/pdf/2609.36301","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.36301","doi":"https://doi.org/10.48550/arxiv.2609.36301","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.36301","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.36301","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36301","pdf_url":"https://arxiv.org/pdf/2609.36301","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320309370","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214947677.pdf","grobid_xml":"https://content.openalex.org/works/W7214947677.grobid-xml"},"referenced_works_count":45,"referenced_works":["https://openalex.org/W2581624817","https://openalex.org/W2794325560","https://openalex.org/W2801655600","https://openalex.org/W2908510526","https://openalex.org/W2912924812","https://openalex.org/W2954698171","https://openalex.org/W2979473749","https://openalex.org/W2988394319","https://openalex.org/W3001279689","https://openalex.org/W3013980252","https://openalex.org/W3040573126","https://openalex.org/W3119866685","https://openalex.org/W3168867926","https://openalex.org/W3212449162","https://openalex.org/W4226153346","https://openalex.org/W4308760184","https://openalex.org/W4387321091","https://openalex.org/W4390723197","https://openalex.org/W4391272939","https://openalex.org/W4391801203","https://openalex.org/W4391988255","https://openalex.org/W4392822410","https://openalex.org/W4396815229","https://openalex.org/W4400434490","https://openalex.org/W4401211627","https://openalex.org/W4403573100","https://openalex.org/W4404308452","https://openalex.org/W4404310926","https://openalex.org/W4404573790","https://openalex.org/W4405432939","https://openalex.org/W4405903187","https://openalex.org/W4406755530","https://openalex.org/W4410356599","https://openalex.org/W4410356661","https://openalex.org/W4414939355","https://openalex.org/W4417466279","https://openalex.org/W7075994024","https://openalex.org/W7077044503","https://openalex.org/W7117108127","https://openalex.org/W7126086621","https://openalex.org/W7133193424","https://openalex.org/W7134042529","https://openalex.org/W7134822015","https://openalex.org/W7160986019","https://openalex.org/W7166846580"],"related_works":["https://openalex.org/W7126205855","https://openalex.org/W7116972716","https://openalex.org/W7212240133","https://openalex.org/W4402427157","https://openalex.org/W4416447653","https://openalex.org/W7154713303","https://openalex.org/W7213936625","https://openalex.org/W4415964461","https://openalex.org/W4405627846","https://openalex.org/W7212384774"],"abstract_inverted_index":{"Mixture-of-Experts":[0],"(MoE)":[1],"layers":[2],"are":[3],"central":[4],"to":[5,68,95],"frontier":[6],"language":[7],"models,":[8],"and":[9,15,31,63,91,112,170],"recent":[10],"architectures":[11],"push":[12],"toward":[13],"more":[14,138],"smaller":[16],"experts.":[17,139],"In":[18],"this":[19,142],"regime,":[20],"the":[21,40,56,65,84,131,167],"standard":[22],"linear":[23],"router":[24,57],"becomes":[25],"a":[26,108,115,134,149],"bottleneck:":[27],"with":[28],"$M$":[29,44,78],"experts":[30,88],"hidden":[32],"dimension":[33],"$h$,":[34],"its":[35],"per-token":[36],"cost":[37,67],"$\u0398(Mh)$":[38],"dominates":[39],"MoE":[41],"layer":[42],"once":[43],"is":[45,89,92],"large.":[46],"We":[47,72,98],"introduce":[48],"MoRE":[49,163],"(Mixture":[50],"of":[51,86,136],"Rank-reduced-routed":[52],"Experts),":[53],"which":[54],"factorizes":[55],"weight":[58],"matrix":[59],"at":[60,153,184],"rank":[61,75,103,122],"$r$":[62],"reduces":[64],"routing":[66,81],"$O((h":[69],"+":[70],"M)r)$.":[71],"prove":[73,100],"that":[74,101,120,155],"logarithmic":[76,102],"in":[77,107,144],"suffices":[79],"for":[80],"expressivity":[82],"when":[83],"number":[85],"active":[87,129],"fixed,":[90],"necessary":[93],"up":[94],"precision":[96],"factors.":[97],"also":[99],"preserves":[104],"load":[105],"balance":[106],"Gaussian":[109],"memorization":[110,165],"model,":[111],"training":[113],"on":[114,160,166,172],"synthetic":[116],"phonebook":[117,168],"task":[118,169],"shows":[119],"low":[121],"does":[123],"not":[124],"hurt":[125],"memorization.":[126],"At":[127],"matched":[128],"FLOPs,":[130],"factorization":[132],"allows":[133],"factor":[135],"$\u0398(h/r)$":[137],"To":[140],"realize":[141],"gain":[143],"wall-clock":[145],"time,":[146],"we":[147],"design":[148],"fused":[150],"Triton":[151],"kernel":[152],"inference":[154],"avoids":[156],"expensive":[157],"memory":[158],"operations":[159],"HBM.":[161],"Empirically,":[162],"improves":[164],"performance":[171],"knowledge-intensive":[173],"Q\\&A":[174],"benchmarks":[175],"after":[176],"pretraining,":[177],"while":[178],"matching":[179],"reasoning":[180],"ability.":[181],"Code":[182],"available":[183],"https://github.com/Matheart/MoRE_code.":[185]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}