{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004173205,0.0001334659,0.0002512282,0.0002178143,0.0002128646,0.0001782502,0.0001266097,0.00009755899,0.000003855322],"category_scores_gemma":[0.0006334082,0.0001455417,0.00003328645,0.0003206366,0.00006817203,0.0002145114,0.00003073951,0.0001407572,0.000001377769],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001616502,"about_ca_system_score_gemma":0.00004898056,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000009128995,"about_ca_topic_score_gemma":0.00007390459,"domain_scores_codex":[0.9988178,0.0001698973,0.0006114869,0.0001767155,0.00009492224,0.0001291709],"domain_scores_gemma":[0.9971226,0.002182196,0.00008882077,0.0003683793,0.0002049506,0.00003299859],"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":[0.00004247329,0.00005285137,0.0006675906,0.0001002723,0.00001782257,9.033563e-8,0.0002087953,0.956761,0.0001011023,0.01976038,0.0001716804,0.02211596],"study_design_scores_gemma":[0.002269138,0.00001822147,0.008777937,0.00005462879,0.00002287209,3.188636e-7,0.0001317508,0.9617521,0.000003873961,0.02551638,0.001329798,0.0001229661],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01717538,0.0002551625,0.9781171,0.0002085113,0.0002834482,0.001214976,0.00003736173,0.0001907021,0.002517356],"genre_scores_gemma":[0.9774316,0.00001824489,0.02194555,0.00005315429,0.00001263646,0.0002912082,0.0001517269,0.00001483233,0.000081051],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9602562,"threshold_uncertainty_score":0.5935019,"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."}}