{"id":"https://openalex.org/W7204576841","doi":"https://doi.org/10.18653/2026.findings-emnlp.1220","title":"Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs","display_name":"Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7204576841","doi":"https://doi.org/10.18653/2026.findings-emnlp.1220"},"language":"en","primary_location":{"id":"doi:10.18653/2026.findings-emnlp.1220","doi":"https://doi.org/10.18653/2026.findings-emnlp.1220","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.1220","pdf_url":"https://aclanthology.org/2026.findings-emnlp.1220.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2026.findings-emnlp.1220.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148484391","display_name":"Rongfeng Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang Rongfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134866346","display_name":"Shichao Weng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shichao Weng","raw_affiliation_strings":["Shenzhen Dobot Corp Ltd , Shenzhen , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Dobot Corp Ltd , Shenzhen , China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133081889","display_name":"Zhiqiang Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhiqiang Wang","raw_affiliation_strings":["Shenzhen Dobot Corp Ltd , Shenzhen , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Dobot Corp Ltd , Shenzhen , China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148541077","display_name":"Xinyu Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xinyu Liu","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114488887","display_name":"Yang Yi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang Yi","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137965391","display_name":"Peilong Zhou","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peilong Zhou","raw_affiliation_strings":["Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China","School of Advanced Interdisciplinary Sciences , University of Chinese Academy of Sciences , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology , Chinese Academy of Sciences , Beijing , China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"raw_affiliation_string":"School of Advanced Interdisciplinary Sciences , University of Chinese Academy of Sciences , Beijing , China","institution_ids":["https://openalex.org/I4210165038"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5023195494","display_name":"Hongwei Tang","orcid":"https://orcid.org/0000-0002-2053-788X","observed_orcids":["https://orcid.org/0000-0002-2053-788X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang Hongwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.70119983,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"25830","last_page":"25842"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.7401999831199646,"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":0.7401999831199646,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.1761000007390976,"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/T10028","display_name":"Topic Modeling","score":0.06069999933242798,"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":0.9620000123977661},{"id":"https://openalex.org/keywords/meta-learning","display_name":"meta-learning","score":0.6959999799728394},{"id":"https://openalex.org/keywords/sparse-routing","display_name":"sparse routing","score":0.3659999966621399},{"id":"https://openalex.org/keywords/metanet","display_name":"METANET","score":0.6230000257492065},{"id":"https://openalex.org/keywords/cross-task-transfer","display_name":"cross-task transfer","score":0.5759999752044678}],"concepts":[{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.5706999897956848},{"id":"https://openalex.org/C2775896111","wikidata":"https://www.wikidata.org/wiki/Q642560","display_name":"Router","level":2,"score":0.5400000214576721},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.5231000185012817},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.47760000824928284},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4708999991416931},{"id":"https://openalex.org/C203479927","wikidata":"https://www.wikidata.org/wiki/Q5165939","display_name":"Controller (irrigation)","level":2,"score":0.4562999904155731},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.42739999294281006},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4221000075340271},{"id":"https://openalex.org/C58328972","wikidata":"https://www.wikidata.org/wiki/Q184609","display_name":"Expert system","level":2,"score":0.40560001134872437},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.3806000053882599},{"id":"https://openalex.org/C131918245","wikidata":"https://www.wikidata.org/wiki/Q1409090","display_name":"Fetch","level":2,"score":0.3352000117301941},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.31060001254081726},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3102000057697296},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.3100999891757965},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.30660000443458557},{"id":"https://openalex.org/C42866809","wikidata":"https://www.wikidata.org/wiki/Q466764","display_name":"Fixed cost","level":2,"score":0.30379998683929443},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2996000051498413},{"id":"https://openalex.org/C164916973","wikidata":"https://www.wikidata.org/wiki/Q1378386","display_name":"Leaky bucket","level":3,"score":0.2816999852657318},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.27079999446868896},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C182365436","wikidata":"https://www.wikidata.org/wiki/Q50701","display_name":"Variable (mathematics)","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.2500999867916107}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/2026.findings-emnlp.1220","doi":"https://doi.org/10.18653/2026.findings-emnlp.1220","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.1220","pdf_url":"https://aclanthology.org/2026.findings-emnlp.1220.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2608.26650","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.26650","pdf_url":"https://arxiv.org/pdf/2608.26650","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.26650","doi":"https://doi.org/10.48550/arxiv.2608.26650","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.26650","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":"doi:10.18653/2026.findings-emnlp.1220","doi":"https://doi.org/10.18653/2026.findings-emnlp.1220","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.1220","pdf_url":"https://aclanthology.org/2026.findings-emnlp.1220.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204576841.pdf","grobid_xml":"https://content.openalex.org/works/W7204576841.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7223247004","https://openalex.org/W7211902890","https://openalex.org/W7223096217","https://openalex.org/W7153671113","https://openalex.org/W7223157615","https://openalex.org/W7203694543","https://openalex.org/W4416035070","https://openalex.org/W4399353875","https://openalex.org/W7171299342","https://openalex.org/W7202186332"],"abstract_inverted_index":{"Mixture-of-Experts":[0],"(MoE)":[1],"models":[2],"route":[3],"each":[4,86],"token":[5],"to":[6,104,141],"a":[7,22,72,81,100,107],"subset":[8],"of":[9,25],"expert":[10,37],"networks,":[11],"increasing":[12],"capacity":[13,41],"while":[14,59,88],"keeping":[15,89],"per-token":[16],"computation":[17],"sparse.Many":[18],"deployed":[19],"MoEs":[20],"use":[21],"fixed":[23,105,153],"number":[24],"active":[26],"experts":[27,112,129,147],"across":[28,57],"layers":[29],"and":[30,39,45,54,80,93],"tasks,":[31,58],"despite":[32],"evidence":[33],"that":[34,75],"layer":[35,87],"roles,":[36],"redundancy,":[38],"required":[40],"vary":[42],"with":[43,117],"depth":[44],"input":[46],"difficulty.Existing":[47],"layer-wise":[48],"methods":[49,61],"usually":[50],"determine":[51],"allocations":[52],"offline":[53],"reuse":[55],"them":[56],"token-level":[60],"rely":[62],"on":[63,113,130,137,148],"local":[64],"routing":[65,83],"signals":[66],"without":[67,143],"task-level":[68],"context.We":[69],"propose":[70],"MetaNet,":[71],"support-set":[73],"controller":[74],"predicts":[76],"an":[77,124],"expertretention":[78],"threshold":[79],"bounded":[82],"bias":[84],"for":[85],"the":[90],"backbone,":[91],"experts,":[92],"original":[94],"router":[95],"frozen.On":[96],"DeepSeek-MoE-16B-Chat,":[97],"MetaNet":[98,139],"provides":[99],"tunable":[101],"accuracy-expert-activation":[102],"trade-off.Relative":[103],"k=6,":[106],"conservative":[108],"setting":[109,126],"activates":[110,127],"3.61":[111],"average":[114,131,149],"(40%":[115],"fewer)":[116,133],"comparable":[118],"MMLU":[119],"accuracy":[120],"(0.489":[121],"versus":[122],"0.474);":[123],"aggressive":[125],"2.28":[128],"(62%":[132],"at":[134,155],"0.438":[135],"accuracy.Trained":[136],"MMLU,":[138],"transfers":[140],"C-Eval":[142],"retraining,":[144],"activating":[145],"2.90":[146],"(52%":[150],"fewer":[151],"than":[152],"k=6)":[154],"0.386":[156],"accuracy.":[157]},"counts_by_year":[],"updated_date":"2026-11-02T09:45:03.842371","created_date":"2026-08-29T00:00:00"}