{"id":"https://openalex.org/W7140085869","doi":"https://doi.org/10.18653/v1/2026.eacl-long.357","title":"Tokenizer-Aware Cross-Lingual Adaptation of Decoder-Only LLMs through Embedding Relearning and Swapping","display_name":"Tokenizer-Aware Cross-Lingual Adaptation of Decoder-Only LLMs through Embedding Relearning and Swapping","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140085869","doi":"https://doi.org/10.18653/v1/2026.eacl-long.357"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.eacl-long.357","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.eacl-long.357","pdf_url":"https://aclanthology.org/2026.eacl-long.357.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":"Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.eacl-long.357.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130365065","display_name":"Fan Jiang","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Fan Jiang","raw_affiliation_strings":["The University of Melbourne  Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Melbourne  Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008847059","display_name":"Hong-Lin Yu","orcid":"https://orcid.org/0000-0003-4287-0365"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Honglin Yu","raw_affiliation_strings":["The University of Melbourne  Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Melbourne  Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111813119","display_name":"Grace Chung","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Grace Y Chung","raw_affiliation_strings":["The University of Melbourne  Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Melbourne  Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5130386180","display_name":"Trevor Cohn","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Trevor Cohn","raw_affiliation_strings":["The University of Melbourne  Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Melbourne  Google","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1291425158"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.24291756,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"7606","last_page":"7636"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2378000020980835,"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":0.2378000020980835,"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":0.15449999272823334,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.05559999868273735,"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/adaptation","display_name":"Adaptation (eye)","score":0.538100004196167},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.4593999981880188},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.357699990272522},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.27090001106262207}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5820000171661377},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.538100004196167},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4693000018596649},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.4593999981880188},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.357699990272522},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.27090001106262207},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2484000027179718},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.24070000648498535},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.23669999837875366},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.23659999668598175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.eacl-long.357","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.eacl-long.357","pdf_url":"https://aclanthology.org/2026.eacl-long.357.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":"Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.eacl-long.357","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.eacl-long.357","pdf_url":"https://aclanthology.org/2026.eacl-long.357.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":"Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140085869.pdf","grobid_xml":"https://content.openalex.org/works/W7140085869.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Extending":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"to":[5,58,76,83,104,131,159,172],"new":[6,35],"languages":[7,96,115],"is":[8,48],"challenging,":[9],"with":[10,136],"most":[11],"methods":[12],"proposed":[13],"suffering":[14],"from":[15],"high":[16,67],"computational":[17],"cost":[18],"and":[19,37,51,65,108,119,145],"catastrophic":[20,147],"forgetting":[21],"of":[22,180],"original":[23],"model":[24,42],"capabilities.Embedding":[25],"relearning":[26,109,125],"(Artetxe":[27],"et":[28],"al.,":[29],"2020),":[30],"a":[31,99,105],"technique":[32],"that":[33,161],"creates":[34],"tokenizers":[36,182],"tunes":[38],"embeddings":[39],"on":[40,80,163],"fixed":[41],"weights":[43],"for":[44,60,66,189],"target":[45],"language":[46,92,186],"adaptation,":[47],"both":[49,117],"light-weight":[50],"performant.However,":[52],"it":[53],"has":[54],"only":[55],"been":[56],"shown":[57],"work":[59],"older":[61],"generation":[62,120],"encoder-only":[63],"models":[64,128],"resource":[68],"languages.In":[69],"this":[70,74],"paper,":[71],"we":[72,122],"extend":[73],"framework":[75],"decoder-only":[77],"LLMs":[78],"focusing":[79],"joint":[81],"adaptation":[82],"many":[84],"languages,":[85],"including":[86],"low-resource":[87],"ones.We":[88],"experiment":[89],"in":[90,153,183],"three":[91],"groups":[93],"over":[94],"100":[95],"each.We":[97],"adapt":[98],"pre-trained":[100],"LLM":[101],"via":[102],"switching":[103],"customized":[106],"tokenizer,":[107],"the":[110,155,177],"embedding":[111,124],"layer.Across":[112],"96":[113],"diverse":[114],"spanning":[116],"classification":[118],"tasks,":[121],"show":[123],"improves":[126],"Gemma2":[127],"by":[129],"up":[130],"20%,":[132],"being":[133],"highly":[134],"competitive":[135],"full-weight":[137],"updating":[138],"baselines":[139],"while":[140],"vastly":[141],"more":[142],"computationally":[143],"efficient":[144],"mitigating":[146],"forgetting.This":[148],"translates":[149],"into":[150],"better":[151],"results":[152],"transferring":[154],"improved":[156],"multilingual":[157,168],"performance":[158],"tasks":[160],"build":[162],"core":[164],"English":[165],"abilities":[166],"(e.g.,":[167],"math":[169],"reasoning),":[170],"compared":[171],"various":[173],"baselines.Further":[174],"analysis":[175],"reveals":[176],"critical":[178],"role":[179],"customizing":[181],"achieving":[184],"effective":[185],"transfer,":[187],"particularly":[188],"non-Latin":[190],"script":[191],"languages.":[192]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-03-24T00:00:00"}