{"id":"https://openalex.org/W4405627846","doi":"https://doi.org/10.48550/arxiv.2412.14711","title":"ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing","display_name":"ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing","publication_year":2024,"publication_date":"2024-12-19","ids":{"openalex":"https://openalex.org/W4405627846","doi":"https://doi.org/10.48550/arxiv.2412.14711"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2412.14711","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2412.14711","pdf_url":"https://arxiv.org/pdf/2412.14711","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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/2412.14711","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5115561203","display_name":"Ziteng Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Wang, Ziteng","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Zhu, Jun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Jun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100641283","display_name":"Jianfei Chen","orcid":"https://orcid.org/0000-0003-4107-5706"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jianfei","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5115561203"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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/T11704","display_name":"Mobile Crowdsensing and Crowdsourcing","score":0.9922999739646912,"subfield":{"id":"https://openalex.org/subfields/1706","display_name":"Computer Science Applications"},"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/T11704","display_name":"Mobile Crowdsensing and Crowdsourcing","score":0.9922999739646912,"subfield":{"id":"https://openalex.org/subfields/1706","display_name":"Computer Science Applications"},"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/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.9768999814987183,"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/T12879","display_name":"Distributed Sensor Networks and Detection Algorithms","score":0.9726999998092651,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/differentiable-function","display_name":"Differentiable function","score":0.7325139045715332},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4376988708972931},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.3217369318008423},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.23589187860488892},{"id":"https://openalex.org/keywords/pure-mathematics","display_name":"Pure mathematics","score":0.08300957083702087}],"concepts":[{"id":"https://openalex.org/C202615002","wikidata":"https://www.wikidata.org/wiki/Q783507","display_name":"Differentiable function","level":2,"score":0.7325139045715332},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4376988708972931},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.3217369318008423},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.23589187860488892},{"id":"https://openalex.org/C202444582","wikidata":"https://www.wikidata.org/wiki/Q837863","display_name":"Pure mathematics","level":1,"score":0.08300957083702087}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2412.14711","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2412.14711","pdf_url":"https://arxiv.org/pdf/2412.14711","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2412.14711","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2412.14711","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2412.14711","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2412.14711","pdf_url":"https://arxiv.org/pdf/2412.14711","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320322392","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549"}],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4405627846.pdf"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4387497383","https://openalex.org/W2948807893","https://openalex.org/W2899084033","https://openalex.org/W2778153218","https://openalex.org/W2748952813","https://openalex.org/W1531601525","https://openalex.org/W4391375266","https://openalex.org/W2078814861","https://openalex.org/W2527526854","https://openalex.org/W1976181487"],"abstract_inverted_index":{"Sparsely":[0],"activated":[1],"Mixture-of-Experts":[2],"(MoE)":[3],"models":[4],"are":[5,22],"widely":[6],"adopted":[7],"to":[8,69,125],"scale":[9],"up":[10],"model":[11,110],"capacity":[12],"without":[13],"increasing":[14],"the":[15,55,62,71,76,126],"computation":[16,88],"budget.":[17],"However,":[18],"vanilla":[19,105],"TopK":[20],"routers":[21],"trained":[23],"in":[24],"a":[25,41,48],"discontinuous,":[26],"non-differentiable":[27],"way,":[28],"limiting":[29],"their":[30],"performance":[31],"and":[32,91,114],"scalability.":[33],"To":[34],"address":[35],"this":[36],"issue,":[37],"we":[38],"propose":[39,67],"ReMoE,":[40],"fully":[42],"differentiable":[43],"MoE":[44,107,132],"architecture":[45],"that":[46,101],"offers":[47],"simple":[49],"yet":[50],"effective":[51],"drop-in":[52],"replacement":[53],"for":[54],"conventional":[56],"TopK+Softmax":[57],"routing,":[58],"utilizing":[59],"ReLU":[60],"as":[61],"router":[63],"instead.":[64],"We":[65],"further":[66],"methods":[68],"regulate":[70],"router's":[72],"sparsity":[73],"while":[74,93],"balancing":[75],"load":[77],"among":[78],"experts.":[79],"ReMoE's":[80],"continuous":[81],"nature":[82],"enables":[83],"efficient":[84],"dynamic":[85],"allocation":[86],"of":[87,116,128],"across":[89,108],"tokens":[90],"layers,":[92],"also":[94],"exhibiting":[95],"domain":[96],"specialization.":[97],"Our":[98],"experiments":[99],"demonstrate":[100],"ReMoE":[102,119],"consistently":[103],"outperforms":[104],"TopK-routed":[106],"various":[109],"sizes,":[111],"expert":[112],"counts,":[113],"levels":[115],"granularity.":[117],"Furthermore,":[118],"exhibits":[120],"superior":[121],"scalability":[122],"with":[123],"respect":[124],"number":[127],"experts,":[129],"surpassing":[130],"traditional":[131],"architectures.":[133],"The":[134],"implementation":[135],"based":[136],"on":[137],"Megatron-LM":[138],"is":[139],"available":[140],"at":[141],"https://github.com/thu-ml/ReMoE.":[142]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-10T00:00:00"}