{"id":"https://openalex.org/W4412886693","doi":"https://doi.org/10.18653/v1/2025.acl-long.207","title":"TokAlign: Efficient Vocabulary Adaptation via Token Alignment","display_name":"TokAlign: Efficient Vocabulary Adaptation via Token Alignment","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412886693","doi":"https://doi.org/10.18653/v1/2025.acl-long.207"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.207","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.207","pdf_url":"https://aclanthology.org/2025.acl-long.207.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 63rd Annual Meeting 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/2025.acl-long.207.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5084711248","display_name":"C.\u2010J. LI","orcid":"https://orcid.org/0009-0003-8023-018X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chong Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100319572","display_name":"Jiajun Zhang","orcid":"https://orcid.org/0000-0001-5293-7434"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiajun Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5015785439","display_name":"Chengqing Zong","orcid":"https://orcid.org/0000-0002-9864-3818"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chengqing Zong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4109","last_page":"4126"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","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"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9843999743461609,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9735999703407288,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.8428153991699219},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7858151197433472},{"id":"https://openalex.org/keywords/adaptation","display_name":"Adaptation (eye)","score":0.7162763476371765},{"id":"https://openalex.org/keywords/vocabulary","display_name":"Vocabulary","score":0.5908172130584717},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.24981454014778137},{"id":"https://openalex.org/keywords/linguistics","display_name":"Linguistics","score":0.08278641104698181},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.06306368112564087},{"id":"https://openalex.org/keywords/neuroscience","display_name":"Neuroscience","score":0.05174282193183899}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8428153991699219},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7858151197433472},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.7162763476371765},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.5908172130584717},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.24981454014778137},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.08278641104698181},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.06306368112564087},{"id":"https://openalex.org/C169760540","wikidata":"https://www.wikidata.org/wiki/Q207011","display_name":"Neuroscience","level":1,"score":0.05174282193183899},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.207","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.207","pdf_url":"https://aclanthology.org/2025.acl-long.207.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.207","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.207","pdf_url":"https://aclanthology.org/2025.acl-long.207.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.46000000834465027,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320321133","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412886693.pdf","grobid_xml":"https://content.openalex.org/works/W4412886693.grobid-xml"},"referenced_works_count":1,"referenced_works":["https://openalex.org/W4281758439"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W4388335561","https://openalex.org/W2970530566","https://openalex.org/W4288261899","https://openalex.org/W4307309205","https://openalex.org/W2967478618","https://openalex.org/W4385009901","https://openalex.org/W4385572700"],"abstract_inverted_index":{"Tokenization":[0],"serves":[1],"as":[2,147,149],"a":[3,85],"foundational":[4],"step":[5],"for":[6,89,100,114],"Large":[7],"Language":[8],"Models":[9],"(LLMs)":[10],"to":[11,55,79,126,152],"process":[12],"text.In":[13],"new":[14,102],"domains":[15],"or":[16],"languages,":[17],"the":[18,21,26,57,62,69,76,80,101,117,139,154,157,173],"inefficiency":[19],"of":[20,30,59,122,143,156],"tokenizer":[22],"will":[23],"slow":[24],"down":[25],"training":[27],"and":[28,66,97,111,141],"generation":[29],"LLM.The":[31],"mismatch":[32],"in":[33],"vocabulary":[34,58,78,112],"also":[35],"hinders":[36],"deep":[37],"knowledge":[38,71],"transfer":[39,68],"between":[40,72,162],"LLMs":[41],"like":[42],"token-level":[43,70,164],"distillation.To":[44],"mitigate":[45],"this":[46],"gap,":[47],"we":[48],"propose":[49],"an":[50],"efficient":[51],"method":[52,104],"named":[53],"TokAlign":[54],"replace":[56],"LLM":[60],"from":[61,119],"token":[63,90],"co-occurrences":[64],"view,":[65],"further":[67],"models.It":[73],"first":[74],"aligns":[75],"source":[77],"target":[81],"one":[82],"by":[83],"learning":[84],"oneto-one":[86],"mapping":[87],"matrix":[88],"IDs.Model":[91],"parameters,":[92],"including":[93],"embeddings,":[94],"are":[95],"rearranged":[96],"progressively":[98],"fine-tuned":[99],"vocabulary.Our":[103],"significantly":[105],"improves":[106],"multilingual":[107],"text":[108],"compression":[109],"rates":[110],"initialization":[113],"LLMs,":[115,163],"decreasing":[116],"perplexity":[118],"3.4e":[120],"2":[121,128],"strong":[123],"baseline":[124],"methods":[125],"1.2e":[127],"after":[129],"initialization.Experimental":[130],"results":[131],"on":[132],"models":[133],"across":[134],"multiple":[135],"parameter":[136],"scales":[137],"demonstrate":[138],"effectiveness":[140],"generalization":[142],"TokAlign,":[144],"which":[145],"costs":[146],"few":[148],"5k":[150],"steps":[151],"restore":[153],"performance":[155],"vanilla":[158],"model.After":[159],"unifying":[160],"vocabularies":[161],"distillation":[165],"can":[166],"remarkably":[167],"boost":[168],"(+4.4%":[169],"than":[170],"sentence-level":[171],"distillation)":[172],"base":[174],"model,":[175],"costing":[176],"only":[177],"235M":[178],"tokens.":[179]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}