{"id":"https://openalex.org/W4388505077","doi":"https://doi.org/10.1109/taslp.2023.3331080","title":"Adaptive Ensemble Self-Distillation With Consistent Gradients for Fast Inference of Pretrained Language Models","display_name":"Adaptive Ensemble Self-Distillation With Consistent Gradients for Fast Inference of Pretrained Language Models","publication_year":2023,"publication_date":"2023-11-08","ids":{"openalex":"https://openalex.org/W4388505077","doi":"https://doi.org/10.1109/taslp.2023.3331080"},"language":"en","primary_location":{"id":"doi:10.1109/taslp.2023.3331080","is_oa":false,"landing_page_url":"http://dx.doi.org/10.1109/taslp.2023.3331080","pdf_url":null,"source":{"id":"https://openalex.org/S4210169297","display_name":"IEEE/ACM Transactions on Audio Speech and Language Processing","issn_l":"2329-9290","issn":["2329-9290","2329-9304"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE/ACM Transactions on Audio, Speech, and Language Processing","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101539159","display_name":"Jun Kong","orcid":"https://orcid.org/0000-0003-0221-7687"},"institutions":[{"id":"https://openalex.org/I189210763","display_name":"Yunnan University","ror":"https://ror.org/0040axw97","country_code":"CN","type":"education","lineage":["https://openalex.org/I189210763"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jun Kong","raw_affiliation_strings":["School of Information Science and Engineering, Yunnan University, Kunming, China"],"affiliations":[{"raw_affiliation_string":"School of Information Science and Engineering, Yunnan University, Kunming, China","institution_ids":["https://openalex.org/I189210763"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113414416","display_name":"Jin Wang","orcid":"https://orcid.org/0000-0002-8298-4378"},"institutions":[{"id":"https://openalex.org/I189210763","display_name":"Yunnan University","ror":"https://ror.org/0040axw97","country_code":"CN","type":"education","lineage":["https://openalex.org/I189210763"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jin Wang","raw_affiliation_strings":["School of Information Science and Engineering, Yunnan University, Kunming, China"],"affiliations":[{"raw_affiliation_string":"School of Information Science and Engineering, Yunnan University, Kunming, China","institution_ids":["https://openalex.org/I189210763"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100707916","display_name":"Xuejie Zhang","orcid":"https://orcid.org/0000-0002-6591-0916"},"institutions":[{"id":"https://openalex.org/I189210763","display_name":"Yunnan University","ror":"https://ror.org/0040axw97","country_code":"CN","type":"education","lineage":["https://openalex.org/I189210763"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xuejie Zhang","raw_affiliation_strings":["School of Information Science and Engineering, Yunnan University, Kunming, China"],"affiliations":[{"raw_affiliation_string":"School of Information Science and Engineering, Yunnan University, Kunming, China","institution_ids":["https://openalex.org/I189210763"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5101539159"],"corresponding_institution_ids":["https://openalex.org/I189210763"],"apc_list":null,"apc_paid":null,"fwci":0.1746,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.57416929,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":94},"biblio":{"volume":"32","issue":null,"first_page":"430","last_page":"442"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9993000030517578,"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":0.991100013256073,"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/classifier","display_name":"Classifier (UML)","score":0.8108866214752197},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7590295672416687},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.726528525352478},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.6514689922332764},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.6400703191757202},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.5728468894958496},{"id":"https://openalex.org/keywords/random-subspace-method","display_name":"Random subspace method","score":0.46486029028892517},{"id":"https://openalex.org/keywords/cascading-classifiers","display_name":"Cascading classifiers","score":0.43002089858055115},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.35866495966911316}],"concepts":[{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.8108866214752197},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7590295672416687},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.726528525352478},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.6514689922332764},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6400703191757202},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5728468894958496},{"id":"https://openalex.org/C106135958","wikidata":"https://www.wikidata.org/wiki/Q7291993","display_name":"Random subspace method","level":3,"score":0.46486029028892517},{"id":"https://openalex.org/C40651066","wikidata":"https://www.wikidata.org/wiki/Q5048220","display_name":"Cascading classifiers","level":4,"score":0.43002089858055115},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.35866495966911316},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/taslp.2023.3331080","is_oa":false,"landing_page_url":"http://dx.doi.org/10.1109/taslp.2023.3331080","pdf_url":null,"source":{"id":"https://openalex.org/S4210169297","display_name":"IEEE/ACM Transactions on Audio Speech and Language Processing","issn_l":"2329-9290","issn":["2329-9290","2329-9304"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE/ACM Transactions on Audio, Speech, and Language Processing","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.44999998807907104}],"awards":[{"id":"https://openalex.org/G3061438829","display_name":null,"funder_award_id":"61966038","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":60,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2896457183","https://openalex.org/W2923014074","https://openalex.org/W2924902521","https://openalex.org/W2948947170","https://openalex.org/W2962677625","https://openalex.org/W2962739339","https://openalex.org/W2963748441","https://openalex.org/W2963846996","https://openalex.org/W2965373594","https://openalex.org/W2970120757","https://openalex.org/W2970454332","https://openalex.org/W2971036683","https://openalex.org/W2972324944","https://openalex.org/W2975381464","https://openalex.org/W2978017171","https://openalex.org/W2979826702","https://openalex.org/W2996428491","https://openalex.org/W2998183051","https://openalex.org/W3003048870","https://openalex.org/W3015982254","https://openalex.org/W3034292689","https://openalex.org/W3034457371","https://openalex.org/W3034999214","https://openalex.org/W3035030897","https://openalex.org/W3035038672","https://openalex.org/W3035451444","https://openalex.org/W3038012435","https://openalex.org/W3098576111","https://openalex.org/W3101248447","https://openalex.org/W3104738015","https://openalex.org/W3105966348","https://openalex.org/W3106070274","https://openalex.org/W3160106041","https://openalex.org/W3165511581","https://openalex.org/W3174657338","https://openalex.org/W3177265267","https://openalex.org/W3197726142","https://openalex.org/W3207607259","https://openalex.org/W4224075564","https://openalex.org/W4226003804","https://openalex.org/W4288089799","https://openalex.org/W4385245566","https://openalex.org/W6605323724","https://openalex.org/W6638523607","https://openalex.org/W6760732026","https://openalex.org/W6762100871","https://openalex.org/W6762945437","https://openalex.org/W6763701032","https://openalex.org/W6766673545","https://openalex.org/W6768021236","https://openalex.org/W6768080748","https://openalex.org/W6768851824","https://openalex.org/W6769627184","https://openalex.org/W6775432837","https://openalex.org/W6775706467","https://openalex.org/W6777017071","https://openalex.org/W6778883912","https://openalex.org/W6779313456","https://openalex.org/W6796093880"],"related_works":["https://openalex.org/W2388864896","https://openalex.org/W2407804800","https://openalex.org/W2117265009","https://openalex.org/W2913388591","https://openalex.org/W1995521197","https://openalex.org/W2953702789","https://openalex.org/W2125266525","https://openalex.org/W108062576","https://openalex.org/W354631666","https://openalex.org/W2912744730"],"abstract_inverted_index":{"Conditional":[0],"computation":[1],"algorithms,":[2],"e.g.,":[3],"the":[4,11,24,30,47,50,59,77,91,95,103,111,117,122,125,143,155,161,170,190,194,215,222],"early":[5,132],"exiting":[6,19,133],"(EE)":[7],"strategy,":[8],"can":[9,41],"accelerate":[10],"inference":[12,32,144,204,232],"of":[13,33,49,64,76,124,146,167,193],"pretrained":[14],"language":[15],"models":[16,220],"(PLMs)":[17],"by":[18],"shallow":[20],"layers":[21],"without":[22],"calculating":[23],"entire":[25],"model.":[26],"In":[27],"addition":[28],"to":[29,45,53,120,189,205],"adaptive":[31,136,182,203],"EE":[34,43,65,81,99,104,126,186,195],"prediction":[35],"for":[36,142,176,184,202],"downstream":[37],"tasks,":[38],"self-distillation":[39,138],"(SD)":[40],"encourage":[42],"classifiers":[44,66,82,100,187],"mimic":[46],"behavior":[48],"final":[51,96,112,171],"classifier":[52,97,113,172],"enhance":[54,121],"their":[55],"representation":[56],"capacity.":[57],"However,":[58],"gradients":[60,141],"from":[61,87],"different":[62,181,185],"tasks":[63],"will":[67,83,107],"conflict":[68],"and":[69,79,98,139,208,226,231],"cancel":[70],"one":[71],"another":[72],"out.":[73],"The":[74,197],"parameters":[75],"backbone":[78],"some":[80],"be":[84,116],"implicitly":[85],"prevented":[86],"updating.":[88],"Moreover,":[89],"if":[90],"semantic":[92],"gap":[93],"between":[94],"is":[101],"significant,":[102],"classifier's":[105],"performance":[106,123,230],"decrease.":[108],"That":[109],"is,":[110],"would":[114],"not":[115],"best":[118],"choice":[119],"classifier.":[127],"This":[128],"study":[129],"proposed":[130,216],"an":[131],"strategy":[134],"with":[135,221],"ensemble":[137],"consistent":[140],"acceleration":[145],"PLMs.":[147],"To":[148],"mitigate":[149],"gradient":[150,159],"conflicts,":[151],"we":[152,178],"orthogonally":[153],"projected":[154],"distillation":[156],"loss's":[157,163],"backpropagated":[158],"onto":[160],"classification":[162],"normal":[164],"plane.":[165],"Instead":[166],"directly":[168],"using":[169],"as":[173],"a":[174],"single-teacher":[175],"self-distillation,":[177],"dynamically":[179],"assemble":[180],"teachers":[183],"according":[188],"learning":[191],"abilities":[192],"classifiers.":[196],"accumulative":[198],"decision":[199],"was":[200],"drawn":[201],"make":[206],"accurate":[207],"reliable":[209],"predictions.":[210],"Experimental":[211],"results":[212],"show":[213],"that":[214],"model":[217,229],"outperforms":[218],"existing":[219],"same":[223],"speed-up":[224],"ratio":[225],"effectively":[227],"balances":[228],"time.":[233]},"counts_by_year":[{"year":2024,"cited_by_count":1}],"updated_date":"2025-12-25T23:11:45.687758","created_date":"2025-10-10T00:00:00"}