{"id":"https://openalex.org/W2979636403","doi":"https://doi.org/10.5281/zenodo.3525483","title":"Transformers without Tears: Improving the Normalization of Self-Attention","display_name":"Transformers without Tears: Improving the Normalization of Self-Attention","publication_year":2019,"publication_date":"2019-10-14","ids":{"openalex":"https://openalex.org/W2979636403","doi":"https://doi.org/10.5281/zenodo.3525483","mag":"2979636403"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:1910.05895","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1910.05895","pdf_url":"https://arxiv.org/pdf/1910.05895","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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/1910.05895","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102874067","display_name":"Toan Quang Nguyen","orcid":"https://orcid.org/0000-0003-2734-0622","observed_orcids":["https://orcid.org/0000-0003-2734-0622"]},"institutions":[{"id":"https://openalex.org/I107639228","display_name":"University of Notre Dame","ror":"https://ror.org/00mkhxb43","country_code":"US","type":"education","lineage":["https://openalex.org/I107639228"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nguyen, Toan Q.","raw_affiliation_strings":["University of Notre Dame"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Notre Dame","institution_ids":["https://openalex.org/I107639228"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5037283647","display_name":"Juli\u00e1n Salazar","orcid":"https://orcid.org/0000-0003-0943-7594","observed_orcids":["https://orcid.org/0000-0003-0943-7594"]},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Salazar, Julian","raw_affiliation_strings":["Amazon"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon","institution_ids":["https://openalex.org/I1311688040"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":124,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.7148000001907349,"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":0.7148000001907349,"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":0.23659999668598175,"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":0.01080000028014183,"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/transformer-training","display_name":"transformer training","score":0.5699999928474426},{"id":"https://openalex.org/keywords/l2-normalization","display_name":"L2 normalization","score":0.7839999794960022},{"id":"https://openalex.org/keywords/low-resource-language-translation","display_name":"low-resource translation","score":0.8669999837875366},{"id":"https://openalex.org/keywords/english-vietnamese-translation","display_name":"English-Vietnamese translation","score":0.8640000224113464},{"id":"https://openalex.org/keywords/english-german-translation","display_name":"English-German translation","score":0.7080000042915344},{"id":"https://openalex.org/keywords/gradient-norm","display_name":"gradient norm","score":0.47999998927116394}],"concepts":[{"id":"https://openalex.org/C136886441","wikidata":"https://www.wikidata.org/wiki/Q926129","display_name":"Normalization (sociology)","level":2,"score":0.7364311218261719},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7226120233535767},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6629963517189026},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.5950162410736084},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5513906478881836},{"id":"https://openalex.org/C27158222","wikidata":"https://www.wikidata.org/wiki/Q5532422","display_name":"Generalizability theory","level":2,"score":0.5471082329750061},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5340918898582458},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4936259984970093},{"id":"https://openalex.org/C203005215","wikidata":"https://www.wikidata.org/wiki/Q79798","display_name":"Machine translation","level":2,"score":0.4657752215862274},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.4140055179595947},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.371051162481308},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.36730706691741943},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2560786008834839},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.19079867005348206},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.15663301944732666},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.11270907521247864},{"id":"https://openalex.org/C144024400","wikidata":"https://www.wikidata.org/wiki/Q21201","display_name":"Sociology","level":0,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C19165224","wikidata":"https://www.wikidata.org/wiki/Q23404","display_name":"Anthropology","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":4,"locations":[{"id":"pmh:oai:arXiv.org:1910.05895","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1910.05895","pdf_url":"https://arxiv.org/pdf/1910.05895","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.1910.05895","doi":"https://doi.org/10.48550/arxiv.1910.05895","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.1910.05895","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":"article-journal"},{"id":"doi:10.5281/zenodo.3525483","doi":"https://doi.org/10.5281/zenodo.3525483","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.3525483","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":"article-journal"},{"id":"doi:10.5281/zenodo.3525484","doi":"https://doi.org/10.5281/zenodo.3525484","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.3525484","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":"article-journal"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:1910.05895","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1910.05895","pdf_url":"https://arxiv.org/pdf/1910.05895","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"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":37,"referenced_works":["https://openalex.org/W222053410","https://openalex.org/W1533861849","https://openalex.org/W1815076433","https://openalex.org/W1836465849","https://openalex.org/W2101105183","https://openalex.org/W2183341477","https://openalex.org/W2194775991","https://openalex.org/W2302255633","https://openalex.org/W2514713644","https://openalex.org/W2581377246","https://openalex.org/W2594833348","https://openalex.org/W2626778328","https://openalex.org/W2764124976","https://openalex.org/W2796108585","https://openalex.org/W2806311723","https://openalex.org/W2885724687","https://openalex.org/W2887920589","https://openalex.org/W2888519496","https://openalex.org/W2892009249","https://openalex.org/W2905927205","https://openalex.org/W2911291251","https://openalex.org/W2912555327","https://openalex.org/W2919290281","https://openalex.org/W2933138175","https://openalex.org/W2955227499","https://openalex.org/W2962761235","https://openalex.org/W2962784628","https://openalex.org/W2962801832","https://openalex.org/W2962911098","https://openalex.org/W2963399222","https://openalex.org/W2963410064","https://openalex.org/W2963418779","https://openalex.org/W2963542740","https://openalex.org/W2964085268","https://openalex.org/W2964121744","https://openalex.org/W2970903692","https://openalex.org/W2994689640"],"related_works":["https://openalex.org/W3205328383","https://openalex.org/W6903536371","https://openalex.org/W7168129461","https://openalex.org/W7133235385","https://openalex.org/W7128725319","https://openalex.org/W7077933107","https://openalex.org/W2971191050","https://openalex.org/W3099096615","https://openalex.org/W7127119520","https://openalex.org/W7126159643"],"abstract_inverted_index":{"We":[0,90],"evaluate":[1],"three":[2],"simple,":[3],"normalization-centric":[4],"changes":[5,70],"to":[6,55],"improve":[7],"Transformer":[8],"training.":[9],"First,":[10],"we":[11,30,47],"show":[12],"that":[13],"pre-norm":[14],"residual":[15],"connections":[16],"(PreNorm)":[17],"and":[18,43,82,99,106,117],"smaller":[19],"initializations":[20],"enable":[21],"warmup-free,":[22],"validation-based":[23],"training":[24,42],"with":[25,34],"large":[26],"learning":[27],"rates.":[28],"Second,":[29],"propose":[31],"$\\ell_2$":[32],"normalization":[33],"a":[35,56,83,100],"single":[36],"scale":[37],"parameter":[38],"(ScaleNorm)":[39],"for":[40],"faster":[41],"better":[44],"performance.":[45,124],"Finally,":[46],"reaffirm":[48],"the":[49,111],"effectiveness":[50],"of":[51],"normalizing":[52],"word":[53],"embeddings":[54],"fixed":[57],"length":[58],"(FixNorm).":[59],"On":[60],"five":[61],"low-resource":[62],"translation":[63],"pairs":[64],"from":[65],"TED":[66],"Talks-based":[67],"corpora,":[68],"these":[69],"always":[71],"converge,":[72],"giving":[73],"an":[74],"average":[75],"+1.1":[76],"BLEU":[77,86],"over":[78],"state-of-the-art":[79],"bilingual":[80],"baselines":[81],"new":[84],"32.8":[85],"on":[87],"IWSLT'15":[88],"English-Vietnamese.":[89],"observe":[91],"sharper":[92],"performance":[93],"curves,":[94],"more":[95],"consistent":[96],"gradient":[97],"norms,":[98],"linear":[101],"relationship":[102],"between":[103],"activation":[104],"scaling":[105],"decoder":[107],"depth.":[108],"Surprisingly,":[109],"in":[110],"high-resource":[112],"setting":[113],"(WMT'14":[114],"English-German),":[115],"ScaleNorm":[116],"FixNorm":[118],"remain":[119],"competitive":[120],"but":[121],"PreNorm":[122],"degrades":[123]},"counts_by_year":[{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":13},{"year":2023,"cited_by_count":19},{"year":2022,"cited_by_count":24},{"year":2021,"cited_by_count":36},{"year":2020,"cited_by_count":23},{"year":2019,"cited_by_count":2}],"updated_date":"2026-10-16T09:59:22.235173","created_date":"2019-10-18T00:00:00"}