{"id":"https://openalex.org/W3174544005","doi":"https://doi.org/10.18653/v1/2021.findings-acl.188","title":"MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers","display_name":"MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers","publication_year":2021,"publication_date":"2021-01-01","ids":{"openalex":"https://openalex.org/W3174544005","doi":"https://doi.org/10.18653/v1/2021.findings-acl.188","mag":"3174544005"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2021.findings-acl.188","doi":"https://doi.org/10.18653/v1/2021.findings-acl.188","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2021.findings-acl.188","pdf_url":"https://aclanthology.org/2021.findings-acl.188.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: ACL-IJCNLP 2021","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2021.findings-acl.188.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100370065","display_name":"Wenhui Wang","orcid":"https://orcid.org/0000-0001-5884-7421","observed_orcids":["https://orcid.org/0000-0001-5884-7421"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wenhui Wang","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5053932343","display_name":"Hangbo Bao","orcid":"https://orcid.org/0000-0002-5773-8781","observed_orcids":["https://orcid.org/0000-0002-5773-8781"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hangbo Bao","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5061624006","display_name":"Shaohan Huang","orcid":"https://orcid.org/0000-0003-4324-6337","observed_orcids":["https://orcid.org/0000-0003-4324-6337"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shaohan Huang","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101751776","display_name":"Li Dong","orcid":"https://orcid.org/0000-0003-2036-1009","observed_orcids":["https://orcid.org/0000-0003-2036-1009"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li Dong","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5014662947","display_name":"Furu Wei","orcid":"https://orcid.org/0000-0002-7810-5852","observed_orcids":["https://orcid.org/0000-0002-7810-5852"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Furu Wei","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I1290206253"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1290206253"],"apc_list":null,"apc_paid":null,"fwci":17.5811,"has_fulltext":true,"cited_by_count":372,"citation_normalized_percentile":{"value":0.99536202,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":88,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"2140","last_page":"2151"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.57669997215271,"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":0.57669997215271,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.38429999351501465,"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":0.023800000548362732,"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/knowledge-distillation","display_name":"knowledge distillation","score":0.7910000085830688},{"id":"https://openalex.org/keywords/multi-head-self-attention","display_name":"multi-head self-attention","score":0.8939999938011169},{"id":"https://openalex.org/keywords/transformer-compression","display_name":"transformer compression","score":0.5239999890327454},{"id":"https://openalex.org/keywords/pre-trained-language-models","display_name":"pre-trained language models","score":0.5839999914169312},{"id":"https://openalex.org/keywords/teacher-student-models","display_name":"teacher-student models","score":0.6520000100135803},{"id":"https://openalex.org/keywords/multilingual-models","display_name":"multilingual models","score":0.6710000038146973},{"id":"https://openalex.org/keywords/bert","display_name":"BERT","score":0.9100000262260437},{"id":"https://openalex.org/keywords/roberta","display_name":"RoBERTa","score":0.9909999966621399},{"id":"https://openalex.org/keywords/xlm-r","display_name":"XLM-R","score":0.9269999861717224},{"id":"https://openalex.org/keywords/minilm","display_name":"MiniLM","score":0.7089999914169312}],"concepts":[{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.6853160858154297},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6723753809928894},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6537322402000427},{"id":"https://openalex.org/C2780312720","wikidata":"https://www.wikidata.org/wiki/Q5689100","display_name":"Head (geology)","level":2,"score":0.4904331862926483},{"id":"https://openalex.org/C25343380","wikidata":"https://www.wikidata.org/wiki/Q277521","display_name":"Relation (database)","level":2,"score":0.42829716205596924},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.1602887511253357},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.15519872307777405},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.12773844599723816},{"id":"https://openalex.org/C127313418","wikidata":"https://www.wikidata.org/wiki/Q1069","display_name":"Geology","level":0,"score":0.10659053921699524},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.10154387354850769},{"id":"https://openalex.org/C43617362","wikidata":"https://www.wikidata.org/wiki/Q170050","display_name":"Chromatography","level":1,"score":0.07121819257736206},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.07020819187164307},{"id":"https://openalex.org/C114793014","wikidata":"https://www.wikidata.org/wiki/Q52109","display_name":"Geomorphology","level":1,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2021.findings-acl.188","doi":"https://doi.org/10.18653/v1/2021.findings-acl.188","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2021.findings-acl.188","pdf_url":"https://aclanthology.org/2021.findings-acl.188.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: ACL-IJCNLP 2021","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2012.15828","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2012.15828","pdf_url":"https://arxiv.org/pdf/2012.15828","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.2012.15828","doi":"https://doi.org/10.48550/arxiv.2012.15828","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2012.15828","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2021.findings-acl.188","doi":"https://doi.org/10.18653/v1/2021.findings-acl.188","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2021.findings-acl.188","pdf_url":"https://aclanthology.org/2021.findings-acl.188.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: ACL-IJCNLP 2021","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W3174544005.pdf","grobid_xml":"https://content.openalex.org/works/W3174544005.grobid-xml"},"referenced_works_count":53,"referenced_works":["https://openalex.org/W131533222","https://openalex.org/W1522301498","https://openalex.org/W1566289585","https://openalex.org/W1599016936","https://openalex.org/W1690739335","https://openalex.org/W1821462560","https://openalex.org/W2130158090","https://openalex.org/W2251939518","https://openalex.org/W2396767181","https://openalex.org/W2525127255","https://openalex.org/W2561238782","https://openalex.org/W2607892599","https://openalex.org/W2626778328","https://openalex.org/W2739351760","https://openalex.org/W2805206884","https://openalex.org/W2806120502","https://openalex.org/W2891555348","https://openalex.org/W2896457183","https://openalex.org/W2923014074","https://openalex.org/W2944815030","https://openalex.org/W2945260553","https://openalex.org/W2950813464","https://openalex.org/W2962753370","https://openalex.org/W2963310665","https://openalex.org/W2963323070","https://openalex.org/W2963341956","https://openalex.org/W2963350559","https://openalex.org/W2963748441","https://openalex.org/W2964121744","https://openalex.org/W2965373594","https://openalex.org/W2969515962","https://openalex.org/W2969601108","https://openalex.org/W2970049541","https://openalex.org/W2970557265","https://openalex.org/W2978017171","https://openalex.org/W2980404057","https://openalex.org/W2981852735","https://openalex.org/W2982399380","https://openalex.org/W2983040767","https://openalex.org/W2986591322","https://openalex.org/W2997666887","https://openalex.org/W2998653236","https://openalex.org/W3007759824","https://openalex.org/W3015233032","https://openalex.org/W3022128899","https://openalex.org/W3082274269","https://openalex.org/W3101045333","https://openalex.org/W3101248447","https://openalex.org/W3101731278","https://openalex.org/W3103884771","https://openalex.org/W3105966348","https://openalex.org/W3171975879","https://openalex.org/W4253067820"],"related_works":["https://openalex.org/W3008374555","https://openalex.org/W3174510164","https://openalex.org/W4393148046","https://openalex.org/W3166068809","https://openalex.org/W7161842911","https://openalex.org/W4402167571","https://openalex.org/W7164298028","https://openalex.org/W4389519284","https://openalex.org/W3166574921","https://openalex.org/W4303447594"],"abstract_inverted_index":{"We":[0],"generalize":[1],"deep":[2],"self-attention":[3,14,43,93],"distillation":[4,16],"in":[5,67,125],"MINILM":[6],"(Wang":[7],"et":[8],"al.,":[9],"2020)":[10],"by":[11,103],"only":[12],"using":[13],"relation":[15],"for":[17,113],"taskagnostic":[18],"compression":[19],"of":[20,35,69,72],"pretrained":[21],"Transformers.In":[22],"particular,":[23],"we":[24,45,106],"define":[25],"multi-head":[26],"selfattention":[27],"relations":[28,94],"as":[29,124],"scaled":[30],"dot-product":[31],"between":[32,87],"the":[33,47,53,70,83,91,99,109,121,155],"pairs":[34],"query,":[36],"key,":[37],"and":[38,58,89,134,147,152],"value":[39],"vectors":[40],"within":[41],"each":[42],"module.Then":[44],"employ":[46],"above":[48],"relational":[49],"knowledge":[50,101],"to":[51,81,96],"train":[52],"student":[54],"model.Besides":[55],"its":[56],"simplicity":[57],"unified":[59],"principle,":[60],"more":[61],"favorably,":[62],"there":[63],"is":[64],"no":[65],"restriction":[66],"terms":[68],"number":[71,86],"student's":[73],"attention":[74],"heads,":[75],"while":[76],"most":[77],"previous":[78],"work":[79],"has":[80],"guarantee":[82],"same":[84],"head":[85],"teacher":[88,114],"student.Moreover,":[90],"fine-grained":[92],"tend":[95],"fully":[97],"exploit":[98],"interaction":[100],"learned":[102],"Transformer.In":[104],"addition,":[105],"thoroughly":[107],"examine":[108],"layer":[110,123],"selection":[111],"strategy":[112],"models,":[115],"rather":[116],"than":[117],"just":[118],"relying":[119],"on":[120,130],"last":[122],"MINILM.We":[126],"conduct":[127],"extensive":[128],"experiments":[129],"compressing":[131],"both":[132],"monolingual":[133],"multilingual":[135],"pre-trained":[136],"models.Experimental":[137],"results":[138],"demonstrate":[139],"that":[140],"our":[141],"models":[142],"1":[143],"distilled":[144],"from":[145],"base-size":[146],"large-size":[148],"teachers":[149],"(BERT,":[150],"RoBERTa":[151],"XLM-R)":[153],"outperform":[154],"state-of-the-art.":[156]},"counts_by_year":[{"year":2026,"cited_by_count":67},{"year":2025,"cited_by_count":117},{"year":2024,"cited_by_count":61},{"year":2023,"cited_by_count":76},{"year":2022,"cited_by_count":39},{"year":2021,"cited_by_count":11},{"year":2020,"cited_by_count":1}],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2021-07-05T00:00:00"}