{"id":"https://openalex.org/W4415306754","doi":"https://doi.org/10.48550/arxiv.2504.17025","title":"Optimizing LLMs for Italian: Reducing Token Fertility and Enhancing Efficiency Through Vocabulary Adaptation","display_name":"Optimizing LLMs for Italian: Reducing Token Fertility and Enhancing Efficiency Through Vocabulary Adaptation","publication_year":2025,"publication_date":"2025-04-23","ids":{"openalex":"https://openalex.org/W4415306754","doi":"https://doi.org/10.48550/arxiv.2504.17025"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2504.17025","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2504.17025","pdf_url":"https://arxiv.org/pdf/2504.17025","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/2504.17025","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114231065","display_name":"Luca Moroni","orcid":"https://orcid.org/0009-0006-1210-5098","observed_orcids":["https://orcid.org/0009-0006-1210-5098"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Moroni, Luca","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045636436","display_name":"Giovanni Puccetti","orcid":"https://orcid.org/0000-0002-8906-0987","observed_orcids":["https://orcid.org/0000-0002-8906-0987","https://orcid.org/0000-0003-1866-5951"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Puccetti, Giovanni","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075199715","display_name":"\u202aPere-Llu\u00eds Huguet Cabot","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cabot, Pere-Lluis Huguet","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5107436210","display_name":"Andrei Stefan Bejgu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bejgu, Andrei Stefan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026996221","display_name":"Edoardo Barba","orcid":"https://orcid.org/0009-0004-4307-1498","observed_orcids":["https://orcid.org/0009-0004-4307-1498"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barba, Edoardo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062667199","display_name":"Alessio Miaschi","orcid":"https://orcid.org/0000-0002-0736-5411","observed_orcids":["https://orcid.org/0000-0002-0736-5411"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Miaschi, Alessio","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084812833","display_name":"Felice Dell\u2019Orletta\u22c4","orcid":"https://orcid.org/0000-0003-3454-9387","observed_orcids":["https://orcid.org/0000-0003-3454-9387"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dell'Orletta, Felice","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082069715","display_name":"Andrea Esuli","orcid":"https://orcid.org/0000-0002-5725-4322","observed_orcids":["https://orcid.org/0000-0002-5725-4322"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Esuli, Andrea","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5026648049","display_name":"Roberto Navigli A","orcid":"https://orcid.org/0000-0003-3831-9706","observed_orcids":["https://orcid.org/0000-0003-3831-9706"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Navigli, Roberto","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":false,"cited_by_count":0,"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/T10028","display_name":"Topic Modeling","score":0.7465999722480774,"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":0.7465999722480774,"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":0.23669999837875366,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.009600000455975533,"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/large-language-models","display_name":"large language models","score":0.6940000057220459},{"id":"https://openalex.org/keywords/italian-language","display_name":"Italian language","score":0.8320000171661377},{"id":"https://openalex.org/keywords/vocabulary-adaptation","display_name":"vocabulary adaptation","score":0.9120000004768372},{"id":"https://openalex.org/keywords/multilingual-llms","display_name":"multilingual LLMs","score":0.6639999747276306},{"id":"https://openalex.org/keywords/parameter-reduction","display_name":"parameter reduction","score":0.5329999923706055},{"id":"https://openalex.org/keywords/continuous-training","display_name":"continuous training","score":0.5950000286102295},{"id":"https://openalex.org/keywords/llama-3-1-8b","display_name":"Llama-3.1-8B","score":0.9049999713897705},{"id":"https://openalex.org/keywords/language-contamination","display_name":"language contamination","score":0.4819999933242798}],"concepts":[{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.763700008392334},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.682200014591217},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6687999963760376},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6111000180244446},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.4973999857902527},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4691999852657318},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.40560001134872437},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3725999891757965},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.36719998717308044},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34380000829696655},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3061000108718872},{"id":"https://openalex.org/C155076860","wikidata":"https://www.wikidata.org/wiki/Q1932004","display_name":"Token economy","level":3,"score":0.3009999990463257},{"id":"https://openalex.org/C518429986","wikidata":"https://www.wikidata.org/wiki/Q964401","display_name":"Fertility","level":3,"score":0.2856000065803528},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.2775999903678894},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.26170000433921814}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2504.17025","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2504.17025","pdf_url":"https://arxiv.org/pdf/2504.17025","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.2504.17025","doi":"https://doi.org/10.48550/arxiv.2504.17025","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2504.17025","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2504.17025","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2504.17025","pdf_url":"https://arxiv.org/pdf/2504.17025","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"number":[1,125],"of":[2,35,65,126,137,151,163],"pretrained":[3],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"is":[8],"increasing":[9],"steadily,":[10],"though":[11],"the":[12,18,74,120,124,135,138,155,161,164],"majority":[13],"are":[14,40],"designed":[15],"predominantly":[16],"for":[17,43,69,73,92],"English":[19,71],"language.":[20,157],"While":[21],"state-of-the-art":[22],"LLMs":[23,72],"can":[24,142],"handle":[25],"other":[26],"languages,":[27,45],"due":[28],"to":[29,47],"language":[30],"contamination":[31],"or":[32],"some":[33],"degree":[34],"multilingual":[36],"pretraining":[37],"data,":[38],"they":[39],"not":[41],"optimized":[42],"non-English":[44],"leading":[46],"inefficient":[48],"encoding":[49],"(high":[50],"token":[51,113],"\"fertility\")":[52],"and":[53,77,117,122,170],"slower":[54],"inference":[55],"speed.":[56],"In":[57],"this":[58],"work,":[59],"we":[60,159],"thoroughly":[61],"compare":[62],"a":[63,85,147],"variety":[64],"vocabulary":[66,93,121],"adaptation":[67,136],"techniques":[68],"optimizing":[70,119],"Italian":[75],"language,":[76],"put":[78],"forward":[79],"Semantic":[80],"Alignment":[81],"Vocabulary":[82],"Adaptation":[83],"(SAVA),":[84],"novel":[86],"method":[87],"that":[88],"leverages":[89],"neural":[90],"mapping":[91],"substitution.":[94],"SAVA":[95],"achieves":[96],"competitive":[97],"performance":[98,145],"across":[99],"multiple":[100],"downstream":[101],"tasks,":[102],"enhancing":[103],"grounded":[104],"alignment":[105],"strategies.":[106],"We":[107,131],"adapt":[108],"two":[109],"LLMs:":[110],"Mistral-7b-v0.1,":[111],"reducing":[112,123],"fertility":[114],"by":[115,128],"25\\%,":[116],"Llama-3.1-8B,":[118],"parameters":[127],"1":[129],"billion.":[130],"show":[132],"that,":[133],"following":[134],"vocabulary,":[139],"these":[140],"models":[141,166],"recover":[143],"their":[144],"with":[146],"relatively":[148],"limited":[149],"stage":[150],"continual":[152],"training":[153],"on":[154,167],"target":[156],"Finally,":[158],"test":[160],"capabilities":[162],"adapted":[165],"various":[168],"multi-choice":[169],"generative":[171],"tasks.":[172]},"counts_by_year":[],"updated_date":"2026-10-14T16:39:01.820549","created_date":"2025-10-10T00:00:00"}