{"id":"W4417167884","doi":"10.48550/arxiv.2504.20216","title":"Flexible extreme thresholds through generalised Bayesian model averaging","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Probability and Risk Models","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Bayesian probability; Sensitivity (control systems); Threshold model; Model selection; Cover (algebra); Term (time); Bayesian inference; Selection (genetic algorithm); Mixture model; Posterior probability","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.003065463,0.0007000106,0.001143026,0.0004080727,0.0005431211,0.0005422115,0.003521887,0.0008283298,0.0005953969],"category_scores_gemma":[0.0008729786,0.0005559748,0.0007524277,0.0008113545,0.0002964839,0.0007296866,0.003467898,0.001398275,0.0003717615],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002372265,"about_ca_system_score_gemma":0.001036827,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004620801,"about_ca_topic_score_gemma":0.0001710509,"domain_scores_codex":[0.9927684,0.0004025052,0.001643386,0.00247601,0.001867446,0.0008422315],"domain_scores_gemma":[0.9942074,0.00053213,0.0005540246,0.003875717,0.0005968642,0.0002338416],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00009098317,0.0002534917,0.1505013,0.0001216716,0.0001350792,0.00002705883,0.004087414,0.8051204,0.0004652953,0.01638766,0.0181531,0.004656593],"study_design_scores_gemma":[0.0003471556,0.00001302792,0.002488255,0.0001514758,0.0000513531,0.000002558347,0.00006173793,0.3695629,0.001491613,0.6219918,0.003284808,0.0005532765],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4061169,0.001154184,0.5523555,0.003772001,0.001699176,0.0006732572,0.0001616798,0.0003446963,0.03372262],"genre_scores_gemma":[0.9486484,0.0003867238,0.01933013,0.002528111,0.0003663409,0.000118503,0.00005136883,0.0000408538,0.02852959],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.6056042,"threshold_uncertainty_score":0.9996892,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.425775411241397,"score_gpt":0.4191508606797159,"score_spread":0.006624550561681097,"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."}}