{"id":"https://openalex.org/W7126376856","doi":"https://doi.org/10.18653/v1/2024.findings-eacl.93","title":"Minimal Distillation Schedule for Extreme Language Model Compression","display_name":"Minimal Distillation Schedule for Extreme Language Model Compression","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W7126376856","doi":"https://doi.org/10.18653/v1/2024.findings-eacl.93"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2024.findings-eacl.93","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.93","pdf_url":"https://aclanthology.org/2024.findings-eacl.93.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2024.findings-eacl.93.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124477213","display_name":"Chen Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124676874","display_name":"Yang Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124568919","display_name":"Qifan Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qifan Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124472917","display_name":"Jiahao Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiahao Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124629799","display_name":"Jingang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jingang Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124562509","display_name":"Wei Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5124471184","display_name":"Dawei Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dawei Song","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"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":0.57223624,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1378","last_page":"1394"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.40290001034736633,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.40290001034736633,"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/T11269","display_name":"Algorithms and Data Compression","score":0.2029000073671341,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.16099999845027924,"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/schedule","display_name":"Schedule","score":0.506600022315979},{"id":"https://openalex.org/keywords/compression","display_name":"Compression (physics)","score":0.5051000118255615},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.4154999852180481},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.33809998631477356},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2759000062942505}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6111000180244446},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.506600022315979},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.5051000118255615},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.4154999852180481},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.33809998631477356},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.31200000643730164},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2870999872684479},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2768000066280365},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2759000062942505},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.25609999895095825}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2024.findings-eacl.93","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.93","pdf_url":"https://aclanthology.org/2024.findings-eacl.93.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.findings-eacl.93","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.93","pdf_url":"https://aclanthology.org/2024.findings-eacl.93.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G4853011787","display_name":null,"funder_award_id":"62376027","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7126376856.pdf","grobid_xml":"https://content.openalex.org/works/W7126376856.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"studies":[1],"have":[2],"revealed":[3],"that":[4,110,133,179],"language":[5,202],"model":[6,100,203],"distillation":[7,33,63,146],"can":[8,155],"become":[9],"less":[10],"effective":[11],"when":[12],"there":[13],"is":[14,116],"a":[15,46,77,129,151,169,201],"significant":[16],"capacity":[17],"gap":[18],"between":[19],"the":[20,23,29,39,42,53,56,69,87,111,114,120,124,135,138,142,148,157,162,174,193],"teacher":[21,31,43,54,61,71,92,125,139,159],"and":[22],"student":[24,115],"models.In":[25],"order":[26],"to":[27,55,67,103,147,187,200],"bridge":[28],"gap,":[30],"assistant-based":[32,62],"has":[34],"been":[35],"introduced,":[36],"in":[37,49,94],"which":[38,85],"selection":[40],"of":[41,89,113,123,137,171,195,206],"assistant":[44,93,140,160],"plays":[45],"crucial":[47],"role":[48],"transferring":[50],"knowledge":[51],"from":[52],"student.However,":[57],"existing":[58],"approaches":[59],"for":[60,98,144],"require":[64],"numerous":[65],"trials":[66],"find":[68],"optimal":[70,91,158],"assistant.In":[72],"this":[73],"paper,":[74],"we":[75,107,191],"propose":[76],"novel":[78],"approach":[79,181],"called":[80],"Minimal":[81],"Distillation":[82],"Schedule":[83],"(MINIDISC),":[84],"enables":[86],"scheduling":[88],"an":[90,183],"just":[95],"one":[96],"trial":[97,145],"extreme":[99],"compression":[101],"(e.g,":[102],"5%":[104],"scale).In":[105],"particular,":[106],"empirically":[108],"show":[109],"performance":[112],"positively":[117],"correlated":[118],"with":[119,161,204],"scale-performance":[121],"tradeoff":[122],"assistant.We":[126],"then":[127],"introduce":[128],"new":[130],"-tradeoff":[131],"metric":[132],"quantifies":[134],"optimality":[136],"without":[141],"need":[143],"student.By":[149],"employing":[150],"sandwich":[152],"framework,":[153],"MINIDISC":[154,167],"select":[156],"best":[163],"tradeoff.We":[164],"extensively":[165],"evaluate":[166],"through":[168],"series":[170],"experiments":[172],"on":[173],"GLUE":[175],"benchmark.The":[176],"results":[177],"demonstrate":[178],"our":[180],"achieved":[182],"improved":[184],"efficiency":[185],"compared":[186],"various":[188],"state-of-the-art":[189],"baselines.Furthermore,":[190],"showcase":[192],"scalability":[194],"MINI-DISC":[196],"by":[197],"applying":[198],"it":[199],"billions":[205],"parameters.":[207],"1":[208]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-02T00:00:00"}