{"id":"W6891677426","doi":"10.48448/34dm-mp79","title":"Do we need Label Regularization to Fine-tune Pre-trained Language Models?","year":2023,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; University of Waterloo","funders":"","keywords":"Regularization (linguistics); Artificial neural network; Language model; Training set; Set (abstract data type); Deep learning","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002750722,0.001269192,0.001053213,0.0003883661,0.0006791542,0.001618496,0.002122684,0.002363396,0.004267478],"category_scores_gemma":[0.01916502,0.0005404663,0.0006894022,0.0005066736,0.001221736,0.004502178,0.001320681,0.00458038,0.003124997],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006569104,"about_ca_system_score_gemma":0.001565816,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0041579,"about_ca_topic_score_gemma":0.009881684,"domain_scores_codex":[0.9986803,0.0004901997,0.00006696762,0.0004366292,0.0002068644,0.0001188963],"domain_scores_gemma":[0.9943409,0.003093386,0.0002918224,0.001551565,0.000511862,0.0002105162],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0008615406,0.0007287907,0.007971972,0.0008568913,0.0004102243,0.0002785044,0.000614421,0.2044692,0.05203121,0.02244367,0.02024632,0.6890873],"study_design_scores_gemma":[0.0001375103,0.0002637428,0.002532135,0.0001849059,0.0001110174,0.0002445768,0.0002581414,0.9122398,0.02897094,0.03732386,0.01766555,0.00006782759],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1087362,0.003244246,0.8616502,0.005787908,0.0005126817,0.0001278704,0.000536876,0.008321038,0.01108291],"genre_scores_gemma":[0.5543885,0.0009825776,0.4317664,0.00237305,0.0002642972,0.000202752,0.00130639,0.002082184,0.006633951],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004267478,"threshold_uncertainty_score":0.01454741,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03342531541763453,"score_gpt":0.3177216516896725,"score_spread":0.284296336272038,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}