{"id":"W4415234144","doi":"10.1080/03610918.2025.2574482","title":"Power-controlled reliability tests for assessing probabilistic classifier calibration","year":2025,"lang":"en","type":"article","venue":"Communications in Statistics - Simulation and Computation","topic":"Fault Detection and Control Systems","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Western University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Probabilistic logic; Calibration; Reliability (semiconductor); Classifier (UML); Pattern recognition (psychology)","routes":{"ca_aff":true,"ca_fund":true,"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":[0.05117597,0.001951635,0.001750254,0.006123083,0.0009810157,0.002202426,0.003124779,0.002656136,0.003670904],"category_scores_gemma":[0.3360746,0.0008313024,0.001628626,0.003946783,0.00358427,0.004430119,0.002629621,0.003717376,0.001074972],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001088599,"about_ca_system_score_gemma":0.001522656,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009840422,"about_ca_topic_score_gemma":0.0006728519,"domain_scores_codex":[0.9524254,0.03059612,0.002668079,0.00452204,0.009204704,0.0005835761],"domain_scores_gemma":[0.6381049,0.2900862,0.02198032,0.02667289,0.02204104,0.001114754],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00171002,0.0004949205,0.04560247,0.001703884,0.001147154,0.0005823042,0.002064604,0.2814298,0.01968554,0.1230898,0.008725597,0.5137639],"study_design_scores_gemma":[0.0002690672,0.001495371,0.0143074,0.000409462,0.0002458431,0.0008674022,0.0003967726,0.8499607,0.02298702,0.1018756,0.00691428,0.0002711319],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.009744488,0.000207105,0.9875714,0.00009465731,0.00004736968,0.0002273348,0.0001249524,0.000770856,0.001211951],"genre_scores_gemma":[0.5312128,0.0002564609,0.4644913,0.0002241727,0.0001356995,0.001743113,0.0005591972,0.0006873109,0.0006898377],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.05117597,"threshold_uncertainty_score":0.2706476,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05952724333346211,"score_gpt":0.3954643187990383,"score_spread":0.3359370754655762,"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."}}