{"id":"W3169323432","doi":"10.48550/arxiv.2106.00099","title":"Multi-Objective SPIBB: Seldonian Offline Policy Improvement with Safety Constraints in Finite MDPs","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Sepsis Diagnosis and Treatment","field":"Medicine","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Reinforcement learning; Baseline (sea); Computer science; Context (archaeology); Task (project management); Bootstrapping (finance); Grid; Mathematical optimization; Operations research; Artificial intelligence; Econometrics; Engineering; Mathematics","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":[0.003777961,0.0015134,0.002111078,0.000713442,0.0004594007,0.001176802,0.001995371,0.002175777,0.003640289],"category_scores_gemma":[0.01295612,0.0008727749,0.0007239901,0.0005691018,0.001947443,0.001697615,0.002167771,0.003080485,0.0005254393],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001677121,"about_ca_system_score_gemma":0.002908164,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005005513,"about_ca_topic_score_gemma":0.003759185,"domain_scores_codex":[0.9985877,0.0006421626,0.00005889318,0.000261546,0.0002684648,0.000181257],"domain_scores_gemma":[0.9905666,0.00756909,0.0005610993,0.0003709844,0.0005789779,0.0003532436],"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":[0.00006854205,0.00004285646,0.0002826838,0.00005127397,0.00001903182,0.00003337276,0.0000354989,0.9816493,0.0002839656,0.005541063,0.0004880715,0.01150421],"study_design_scores_gemma":[0.00001256173,0.00002279484,0.00002276903,0.000005709628,0.000001919217,0.000003918964,0.000002714504,0.9967505,0.0001394109,0.002935092,0.0001007075,0.00000202761],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0239986,0.0003000794,0.9717601,0.0004968829,0.0000381918,0.00009384281,0.00008285145,0.0006768149,0.002552718],"genre_scores_gemma":[0.7655516,0.0002088787,0.2292738,0.0004409929,0.00007566426,0.0004125244,0.00024884,0.0002767322,0.003510928],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.005005513,"threshold_uncertainty_score":0.01998001,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1020663601134205,"score_gpt":0.2481888585012618,"score_spread":0.1461224983878413,"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."}}