{"id":"W3217026627","doi":"10.48550/arxiv.2111.15430","title":"The Devil is in the Margin: Margin-based Label Smoothing for Network Calibration","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Compute Canada","keywords":"Softmax function; Discriminative model; Margin (machine learning); Computer science; Artificial intelligence; Overfitting; Calibration; Machine learning; Artificial neural network; Mathematical optimization; Pattern recognition (psychology); Mathematics; Statistics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001132838,0.0002212607,0.0001981891,0.00009201123,0.0005739338,0.0006962575,0.001752363,0.000178305,0.00001572823],"category_scores_gemma":[0.00008883252,0.000182138,0.0001637205,0.0007853696,0.00007435141,0.0002871319,0.0005020654,0.0005974452,0.000007196233],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000123015,"about_ca_system_score_gemma":0.0003674764,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007652467,"about_ca_topic_score_gemma":0.0002794812,"domain_scores_codex":[0.9981015,0.0004500709,0.0002238418,0.000692438,0.0001424355,0.000389754],"domain_scores_gemma":[0.9977277,0.0009213345,0.0002448702,0.0009200717,0.0001235921,0.0000624189],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00004194944,0.0000556368,0.001751183,0.00006009845,0.00004283624,0.00009130059,0.001736365,0.7714274,0.000006084637,0.2189142,0.001842135,0.004030842],"study_design_scores_gemma":[0.0005292502,0.00002009687,0.001025817,0.00009032413,0.00002027776,0.000001060464,0.0005391953,0.9737688,0.000011426,0.01101979,0.01274685,0.0002271742],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01639102,0.0001995135,0.9779349,0.002989086,0.0006010577,0.0005092285,0.000003732668,0.00008884915,0.001282641],"genre_scores_gemma":[0.9871638,0.00008775864,0.008709027,0.002851591,0.0001210216,0.000008498945,0.00003025488,0.00001748344,0.001010502],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9707729,"threshold_uncertainty_score":0.7427374,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09479604473760554,"score_gpt":0.2060714126245323,"score_spread":0.1112753678869268,"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."}}