{"id":"https://openalex.org/W7220985677","doi":"https://doi.org/10.48550/arxiv.2610.08879","title":"Tiny-Scale Chinese BERT Pretraining: A Controlled Comparison of MLM, WWM, and MacBERT Strategies","display_name":"Tiny-Scale Chinese BERT Pretraining: A Controlled Comparison of MLM, WWM, and MacBERT Strategies","publication_year":2026,"publication_date":"2026-10-06","ids":{"openalex":"https://openalex.org/W7220985677","doi":"https://doi.org/10.48550/arxiv.2610.08879"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.08879","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.08879","pdf_url":"https://arxiv.org/pdf/2610.08879","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/2610.08879","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153503033","display_name":"Yiping Bai","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Bai, Yiping","raw_affiliation_strings":["Guangdong Haiqixing Marine Technology Co., Ltd"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong Haiqixing Marine Technology Co., Ltd","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5153503033"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":0.9937000274658203,"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":0.9937000274658203,"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":0.00430000014603138,"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":0.0003000000142492354,"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/pre-training-strategies","display_name":"pre-training strategies","score":0.7319999933242798},{"id":"https://openalex.org/keywords/masked-language-modeling","display_name":"masked language modeling","score":0.9819999933242798},{"id":"https://openalex.org/keywords/macbert","display_name":"MacBERT","score":0.968999981880188},{"id":"https://openalex.org/keywords/perplexity","display_name":"perplexity","score":0.734000027179718},{"id":"https://openalex.org/keywords/synonym-replacement","display_name":"synonym replacement","score":0.5680000185966492},{"id":"https://openalex.org/keywords/chinese-wikipedia","display_name":"Chinese Wikipedia","score":0.5659999847412109},{"id":"https://openalex.org/keywords/intrinsic-evaluation","display_name":"intrinsic evaluation","score":0.39100000262260437}],"concepts":[{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.9805999994277954},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.7577999830245972},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6758000254631042},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.6535000205039978},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6025999784469604},{"id":"https://openalex.org/C173483453","wikidata":"https://www.wikidata.org/wiki/Q1040689","display_name":"Synonym (taxonomy)","level":3,"score":0.5889999866485596},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.4846999943256378},{"id":"https://openalex.org/C2777402240","wikidata":"https://www.wikidata.org/wiki/Q6783436","display_name":"Masking (illustration)","level":2,"score":0.4699999988079071},{"id":"https://openalex.org/C90805587","wikidata":"https://www.wikidata.org/wiki/Q10944557","display_name":"Word (group theory)","level":2,"score":0.44929999113082886},{"id":"https://openalex.org/C2777530160","wikidata":"https://www.wikidata.org/wiki/Q41796","display_name":"Sentence","level":2,"score":0.44690001010894775},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.445499986410141},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.38530001044273376},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.37389999628067017},{"id":"https://openalex.org/C155092808","wikidata":"https://www.wikidata.org/wiki/Q182557","display_name":"Computational linguistics","level":2,"score":0.358599990606308},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2892000079154968},{"id":"https://openalex.org/C3018428822","wikidata":"https://www.wikidata.org/wiki/Q7850","display_name":"Chinese language","level":2,"score":0.27489998936653137},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C35639132","wikidata":"https://www.wikidata.org/wiki/Q7452468","display_name":"Sequence labeling","level":3,"score":0.257099986076355},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2565999925136566},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.25589999556541443}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.08879","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.08879","pdf_url":"https://arxiv.org/pdf/2610.08879","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.2610.08879","doi":"https://doi.org/10.48550/arxiv.2610.08879","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.08879","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":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.08879","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.08879","pdf_url":"https://arxiv.org/pdf/2610.08879","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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7220985677.pdf","grobid_xml":"https://content.openalex.org/works/W7220985677.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2952370363","https://openalex.org/W4221152448","https://openalex.org/W4353007076","https://openalex.org/W3021636956","https://openalex.org/W3197120431","https://openalex.org/W4416037556","https://openalex.org/W4386566638","https://openalex.org/W4223589941","https://openalex.org/W3034394361","https://openalex.org/W3104223418"],"abstract_inverted_index":{"Pretraining":[0],"strategies":[1,41],"significantly":[2],"impact":[3],"the":[4,95,157,175,181],"quality":[5],"of":[6,12,38,102],"language":[7],"models,":[8],"yet":[9,180],"existing":[10],"comparisons":[11],"Masked":[13],"Language":[14],"Modeling":[15],"(MLM),":[16],"Whole":[17],"Word":[18],"Masking":[19],"(WWM),":[20],"and":[21,64,72,87,117,198],"MacBERT-style":[22],"replacement":[23,194],"have":[24],"focused":[25],"primarily":[26],"on":[27,42],"base-scale":[28,159],"models":[29,69,197],"(>=110M":[30],"parameters).":[31,54],"This":[32],"paper":[33],"presents":[34],"a":[35,43,114,127,146,169],"controlled":[36],"comparison":[37],"these":[39],"three":[40,68],"tiny-scale":[44],"Chinese":[45,62],"BERT":[46],"model":[47],"(4":[48],"layers,":[49],"256":[50],"hidden":[51],"dimensions,":[52],"8.7M":[53],"Under":[55],"identical":[56],"architecture,":[57],"corpus":[58,199],"(1.29M":[59],"sentences":[60],"from":[61,70,156],"Wikipedia),":[63],"hyperparameters,":[65],"we":[66],"train":[67],"scratch":[71],"evaluate":[73],"them":[74],"across":[75],"five":[76],"intrinsic":[77,98],"dimensions:":[78],"perplexity,":[79],"MLM":[80,93,118],"hit":[81,119],"rate,":[82],"semantic":[83],"discrimination,":[84],"grammatical":[85],"judgment,":[86],"contextual":[88],"sensitivity.":[89],"At":[90],"tiny":[91],"scale,":[92],"achieves":[94,174],"best":[96],"overall":[97],"performance":[99],"(winning":[100],"3":[101],"5":[103],"dimensions),":[104],"while":[105],"WWM":[106,150,163],"excels":[107],"in":[108],"both":[109],"perplexity":[110,138,183],"(1.27":[111],"vs.":[112,122],"2.10,":[113],"39.5%":[115],"improvement)":[116],"rate":[120],"(22%":[121],"16%).":[123],"Notably,":[124],"MacBERT":[125,173],"under":[126,192],"severely":[128],"limited":[129],"synonym":[130],"dictionary":[131],"(222":[132],"entries,":[133],"3.3%":[134],"coverage)":[135],"exhibits":[136],"severe":[137],"degradation":[139],"(47.23,":[140],"22x":[141],"higher":[142],"than":[143],"MLM),":[144],"yielding":[145],"ranking":[147],"(MLM":[148],">":[149,162,164],">>":[151],"MacBERT)":[152],"that":[153,186],"differs":[154],"markedly":[155],"established":[158],"conclusion":[160],"(MacBERT":[161],"MLM).":[165],"We":[166],"further":[167],"identify":[168],"critical":[170],"evaluation":[171],"pitfall:":[172],"lowest":[176],"training":[177,187],"loss":[178,188],"(2.17)":[179],"highest":[182],"(47.23),":[184],"revealing":[185],"alone":[189],"is":[190],"unreliable":[191],"mixed":[193],"strategies.":[195],"All":[196],"are":[200],"publicly":[201],"available":[202],"at":[203],"https://huggingface.co/eebyp.":[204]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-10-09T00:00:00"}