{"id":"https://openalex.org/W7214404607","doi":"https://doi.org/10.48550/arxiv.2609.29266","title":"Baszta: Data-Centric Fine-Tuning of a Polish Multi-Label Safety Classifier","display_name":"Baszta: Data-Centric Fine-Tuning of a Polish Multi-Label Safety Classifier","publication_year":2026,"publication_date":"2026-09-24","ids":{"openalex":"https://openalex.org/W7214404607","doi":"https://doi.org/10.48550/arxiv.2609.29266"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.29266","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.29266","pdf_url":"https://arxiv.org/pdf/2609.29266","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.29266","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5152955745","display_name":"Adam G\u00f3rski","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"G\u00f3rski, Adam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5152954971","display_name":"Mateusz J\u0105kalak","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"J\u0105kalak, Mateusz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5004393080","display_name":"Rafa\u0142 Jakubowski","orcid":"https://orcid.org/0000-0001-6837-5931","observed_orcids":["https://orcid.org/0000-0001-6837-5931"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jakubowski, Rafa\u0142","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":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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9916999936103821,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9916999936103821,"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/T11550","display_name":"Text and Document Classification Technologies","score":0.004399999976158142,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.0020000000949949026,"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/multi-label-text-classification","display_name":"multi-label text classification","score":0.7789999842643738},{"id":"https://openalex.org/keywords/polish-language","display_name":"Polish language","score":0.8690000176429749},{"id":"https://openalex.org/keywords/fine-tuning","display_name":"fine-tuning","score":0.6579999923706055},{"id":"https://openalex.org/keywords/bert","display_name":"BERT","score":0.4909999966621399},{"id":"https://openalex.org/keywords/focal-loss","display_name":"focal loss","score":0.8619999885559082},{"id":"https://openalex.org/keywords/out-of-distribution-generalization","display_name":"out-of-distribution generalization","score":0.6909999847412109},{"id":"https://openalex.org/keywords/calibration","display_name":"calibration","score":0.7789999842643738},{"id":"https://openalex.org/keywords/temperature-scaling","display_name":"temperature scaling","score":0.8870000243186951},{"id":"https://openalex.org/keywords/platt-scaling","display_name":"Platt scaling","score":0.9390000104904175},{"id":"https://openalex.org/keywords/isotonic-regression","display_name":"isotonic regression","score":0.9480000138282776},{"id":"https://openalex.org/keywords/hate-speech","display_name":"hate speech","score":0.7080000042915344},{"id":"https://openalex.org/keywords/vulgar-language","display_name":"vulgarity","score":0.7649999856948853},{"id":"https://openalex.org/keywords/self-harm","display_name":"self-harm","score":0.7210000157356262},{"id":"https://openalex.org/keywords/mean-pooling","display_name":"mean pooling","score":0.746999979019165},{"id":"https://openalex.org/keywords/adversarial-robustness","display_name":"adversarial robustness","score":0.6819999814033508},{"id":"https://openalex.org/keywords/threshold-adjustment","display_name":"threshold adjustment","score":0.4749999940395355}],"concepts":[{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4984000027179718},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.4645000100135803},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.44369998574256897},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.4375},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.421099990606308},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.40860000252723694},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.38040000200271606},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.37880000472068787},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.36070001125335693},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.35440000891685486},{"id":"https://openalex.org/C89128539","wikidata":"https://www.wikidata.org/wiki/Q1949963","display_name":"Statistic","level":2,"score":0.3172000050544739},{"id":"https://openalex.org/C106347477","wikidata":"https://www.wikidata.org/wiki/Q5384228","display_name":"Equating","level":3,"score":0.3075999915599823},{"id":"https://openalex.org/C17418463","wikidata":"https://www.wikidata.org/wiki/Q3455874","display_name":"Isotonic regression","level":3,"score":0.30660000443458557},{"id":"https://openalex.org/C169903167","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Test set","level":2,"score":0.28600001335144043},{"id":"https://openalex.org/C141141315","wikidata":"https://www.wikidata.org/wiki/Q2379942","display_name":"Guard (computer science)","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.2782000005245209},{"id":"https://openalex.org/C166955791","wikidata":"https://www.wikidata.org/wiki/Q629579","display_name":"Macro","level":2,"score":0.27309998869895935},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.271699994802475},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26649999618530273},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.26499998569488525},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25220000743865967},{"id":"https://openalex.org/C141404830","wikidata":"https://www.wikidata.org/wiki/Q2823869","display_name":"AdaBoost","level":3,"score":0.25060001015663147}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.29266","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.29266","pdf_url":"https://arxiv.org/pdf/2609.29266","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.2609.29266","doi":"https://doi.org/10.48550/arxiv.2609.29266","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.29266","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:2609.29266","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.29266","pdf_url":"https://arxiv.org/pdf/2609.29266","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/W7214404607.pdf","grobid_xml":"https://content.openalex.org/works/W7214404607.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3193471994","https://openalex.org/W4285223045","https://openalex.org/W4385374425","https://openalex.org/W4401306886","https://openalex.org/W4412886759","https://openalex.org/W4416116727","https://openalex.org/W7092296780","https://openalex.org/W7128496732","https://openalex.org/W7166858143"],"related_works":["https://openalex.org/W7128496732","https://openalex.org/W7222999228","https://openalex.org/W7204509777","https://openalex.org/W4412889146","https://openalex.org/W7203620323","https://openalex.org/W7162783036","https://openalex.org/W7167286471","https://openalex.org/W7214565461","https://openalex.org/W7117152586","https://openalex.org/W7171739582"],"abstract_inverted_index":{"We":[0,91,231],"develop":[1],"a":[2,21,61,84,106,133,217,221],"multi-label":[3],"Polish":[4],"content-safety":[5],"classifier":[6,107],"by":[7],"fine-tuning":[8],"allegro/herbert-base-cased":[9],"(124M)":[10],"across":[11],"five":[12],"categories":[13],"(hate,":[14],"vulgarity,":[15],"sexual":[16],"content,":[17],"crime,":[18],"self-harm)":[19],"using":[20],"Focal":[22],"+":[23],"R-Drop":[24],"objective,":[25],"and":[26,114,136,144,171,181,216,251],"evaluate":[27],"the":[28,36,50,123,139,176,194,226,239,261],"resulting":[29],"model":[30,59,86],"against":[31,87],"Bielik":[32],"Guard":[33],"(S\u00f3jka)":[34],"on":[35,49,111,122,193,208,212],"shared":[37],"out-of-distribution":[38,154,262],"Gadzi":[39,101,200],"J\u0119zyk":[40,102,201],"benchmark.":[41],"Both":[42],"systems":[43],"are":[44,147,245],"given":[45],"per-category":[46,172],"threshold":[47],"tuning":[48],"same":[51,124],"calibration":[52,159,195],"split.":[53],"Under":[54],"that":[55,95,108,141,244,266],"matched":[56],"protocol":[57],"our":[58],"holds":[60],"small":[62],"but":[63],"statistically":[64],"significant":[65],"lead":[66,74],"in":[67,149],"micro":[68,96,120,128],"F1,":[69],"while":[70,167,259],"an":[71,80,88],"apparent":[72],"macro-F1":[73],"does":[75],"not":[76],"survive:":[77],"it":[78,209],"was":[79],"artifact":[81],"of":[82,158,228],"comparing":[83],"tuned":[85],"untuned":[89],"one.":[90,241],"also":[92],"report":[93,232],"what":[94],"figure":[97],"is":[98,103,138,156],"worth.":[99],"Because":[100],"97%":[104],"crime-positive,":[105],"flags":[109],"crime":[110,211],"every":[112,213],"input":[113],"nothing":[115],"else":[116],"already":[117],"scores":[118],"0.910":[119],"F1":[121,258],"test":[125],"split,":[126],"so":[127,205],"separates":[129],"neither":[130],"system":[131],"from":[132,277],"degenerate":[134],"strategy":[135],"macro":[137,257],"column":[140],"does.":[142],"Per-category":[143],"per-protocol":[145],"figures":[146],"reported":[148],"Section":[150],"4.":[151],"The":[152],"residual":[153],"gap":[155],"one":[157],"rather":[160,236,274],"than":[161,237,275],"discrimination.":[162],"Ranking":[163],"quality":[164],"stays":[165],"high":[166],"positive":[168],"probabilities":[169],"collapse,":[170],"temperature":[173],"scaling":[174,180],"recovers":[175],"loss":[177],"where":[178],"Platt":[179],"isotonic":[182],"regression":[183],"do":[184],"not.":[185],"That":[186],"recovery":[187],"turns":[188],"out":[189],"to":[190,269],"be":[191,270],"conditional":[192],"set":[196],"containing":[197],"safe":[198,214],"text.":[199],"contains":[202],"almost":[203],"none,":[204],"thresholds":[206],"fitted":[207],"flag":[210],"input,":[215],"balanced":[218],"refit":[219],"buys":[220],"deployable":[222],"operating":[223,234],"point":[224],"at":[225],"cost":[227],"adversarial":[229],"recall.":[230],"both":[233],"points":[235],"only":[238],"flattering":[240],"Two":[242],"changes":[243],"standard":[246],"practice,":[247],"per-class":[248],"cost-sensitive":[249],"weighting":[250],"mean":[252],"pooling,":[253],"each":[254],"raise":[255],"in-distribution":[256,278],"lowering":[260],"figure,":[263],"which":[264],"indicates":[265],"robustness":[267],"has":[268],"selected":[271],"for":[272],"directly":[273],"inherited":[276],"accuracy.":[279]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-26T00:00:00"}