{"id":"https://openalex.org/W7140144174","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.126","title":"Representation Collapse in Machine Translation Through the Lens of Angular Dispersion","display_name":"Representation Collapse in Machine Translation Through the Lens of Angular Dispersion","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140144174","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.126"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.126","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.126","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.126","pdf_url":"https://aclanthology.org/2026.findings-eacl.126.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","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: EACL 2026","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.findings-eacl.126.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5088889495","display_name":"Evgeniia Tokarchuk","orcid":"https://orcid.org/0000-0002-7163-863X","observed_orcids":["https://orcid.org/0000-0002-7163-863X"]},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Evgeniia Tokarchuk","raw_affiliation_strings":["Language Technology Lab , University of Amsterdam"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Language Technology Lab , University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114618394","display_name":"Maya K. Nachesa","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Maya K. Nachesa","raw_affiliation_strings":["Language Technology Lab , University of Amsterdam"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Language Technology Lab , University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108046115","display_name":"Sergey Troshin","orcid":"https://orcid.org/0000-0002-8439-390X","observed_orcids":["https://orcid.org/0000-0002-8439-390X"]},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Sergey Troshin","raw_affiliation_strings":["Language Technology Lab , University of Amsterdam"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Language Technology Lab , University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5064085103","display_name":"Vlad Niculae","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Vlad Niculae","raw_affiliation_strings":["Language Technology Lab , University of Amsterdam"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Language Technology Lab , University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I887064364"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.15270728,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2420","last_page":"2431"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.7689999938011169,"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":0.7689999938011169,"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":0.2020999938249588,"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":0.011300000362098217,"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/representation-collapse","display_name":"representation collapse","score":0.9980000257492065},{"id":"https://openalex.org/keywords/angular-dispersion","display_name":"angular dispersion","score":0.8259999752044678},{"id":"https://openalex.org/keywords/neural-machine-translation","display_name":"neural machine translation","score":0.9419999718666077},{"id":"https://openalex.org/keywords/transformer","display_name":"transformer models","score":0.5139999985694885},{"id":"https://openalex.org/keywords/next-token-prediction","display_name":"next-token prediction","score":0.6800000071525574},{"id":"https://openalex.org/keywords/regularization","display_name":"regularization","score":0.5289999842643738},{"id":"https://openalex.org/keywords/quantized-models","display_name":"quantized models","score":0.7419999837875366}],"concepts":[{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.5218999981880188},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.4927999973297119},{"id":"https://openalex.org/C120665830","wikidata":"https://www.wikidata.org/wiki/Q14620","display_name":"Optics","level":1,"score":0.4839000105857849},{"id":"https://openalex.org/C177562468","wikidata":"https://www.wikidata.org/wiki/Q182893","display_name":"Dispersion (optics)","level":2,"score":0.4011000096797943},{"id":"https://openalex.org/C15336307","wikidata":"https://www.wikidata.org/wiki/Q1766051","display_name":"Lens (geology)","level":2,"score":0.3993000090122223},{"id":"https://openalex.org/C149364088","wikidata":"https://www.wikidata.org/wiki/Q185917","display_name":"Translation (biology)","level":4,"score":0.3930000066757202},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.36329999566078186},{"id":"https://openalex.org/C5339829","wikidata":"https://www.wikidata.org/wiki/Q1425977","display_name":"Machine vision","level":2,"score":0.3131999969482422},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2921000123023987},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.28349998593330383}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.126","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.126","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.126","pdf_url":"https://aclanthology.org/2026.findings-eacl.126.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","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: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.126","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.126","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.126","pdf_url":"https://aclanthology.org/2026.findings-eacl.126.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","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: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[{"id":"https://metadata.un.org/sdg/10","score":0.4427242577075958,"display_name":"Reduced inequalities"}],"study_designs":[],"awards":[{"id":"https://openalex.org/G6640760388","display_name":"Trustworthy AI-based Systems for Sustainable Growth","funder_award_id":"KICH3.LTP.20.006","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"}],"funders":[{"id":"https://openalex.org/F4320321800","display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek","ror":"https://ror.org/04jsz6e67"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140144174.pdf","grobid_xml":"https://content.openalex.org/works/W7140144174.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W3012638462","https://openalex.org/W3035511085","https://openalex.org/W2985943212","https://openalex.org/W2962931466","https://openalex.org/W3114856720","https://openalex.org/W2981910001","https://openalex.org/W3066373881","https://openalex.org/W4281656837","https://openalex.org/W4389518736","https://openalex.org/W3098457691"],"abstract_inverted_index":{"Modern":[0],"neural":[1,77],"translation":[2,135],"models":[3,141],"based":[4,119],"on":[5,18,120],"the":[6,50,53,64,81,91,98,148],"Transformer":[7,55],"architecture":[8],"are":[9,152],"known":[10],"for":[11],"their":[12],"high":[13],"performance,":[14],"particularly":[15],"when":[16],"trained":[17],"high-resource":[19],"datasets.A":[20],"standard":[21],"nexttoken":[22],"prediction":[23],"training":[24,74],"strategy,":[25],"while":[26],"widely":[27],"adopted":[28],"in":[29,49,72],"practice,":[30],"may":[31],"lead":[32],"to":[33,61,86,90],"overlooked":[34],"artifacts":[35],"such":[36],"as":[37],"representation":[38,51,101],"collapse.Previous":[39],"works":[40],"have":[41],"shown":[42],"that":[43,126,139,147],"this":[44,94],"problem":[45],"is":[46,68],"especially":[47],"pronounced":[48],"of":[52,75,100,106,150],"deeper":[54],"layers,":[56],"where":[57,80],"it":[58,127],"often":[59],"fails":[60],"efficiently":[62],"utilize":[63],"geometric":[65],"space.Representation":[66],"collapse":[67,102,131,144],"even":[69,154],"more":[70],"evident":[71],"end-to-end":[73],"continuous-output":[76],"machine":[78],"translation,":[79],"trivial":[82],"solution":[83],"would":[84],"be":[85],"set":[87],"all":[88],"vectors":[89],"same":[92],"value.In":[93],"work,":[95],"we":[96,137],"analyze":[97],"dynamics":[99],"at":[103],"different":[104],"levels":[105],"discrete":[107],"and":[108,123,146],"continuous":[109],"NMT":[110],"transformers":[111],"throughout":[112],"training.We":[113],"incorporate":[114],"an":[115],"existing":[116],"regularization":[117,151],"method":[118],"angular":[121],"dispersion":[122],"demonstrate":[124],"empirically":[125],"not":[128],"only":[129],"mitigates":[130],"but":[132],"also":[133],"improves":[134],"quality.Furthermore,":[136],"show":[138],"quantized":[140],"exhibit":[142],"similar":[143],"behavior":[145],"benefits":[149],"preserved":[153],"after":[155],"quantization.":[156]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-24T00:00:00"}