{"id":"W2972435663","doi":"10.65109/hlnw2204","title":"Safe Policy Improvement with an Estimated Baseline Policy","year":2020,"lang":"en","type":"preprint","venue":"","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada)","funders":"","keywords":"Baseline (sea); Reinforcement learning; Bootstrapping (finance); Computer science; Variance (accounting); Control (management); Machine learning; Artificial intelligence; Econometrics; Mathematics; Economics; Political science; Accounting","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0003838702,0.000568525,0.0004926175,0.0004791488,0.0001348998,0.0007364164,0.002749154,0.0002284325,0.00007820794],"category_scores_gemma":[0.0002793731,0.0004623307,0.0001016636,0.0008629679,0.00007677879,0.0003655237,0.003591252,0.0008427817,0.0001784029],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003764164,"about_ca_system_score_gemma":0.002304096,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003061216,"about_ca_topic_score_gemma":0.00002626452,"domain_scores_codex":[0.996626,0.0001115939,0.0005980319,0.001201361,0.0008037845,0.0006591852],"domain_scores_gemma":[0.9965884,0.00006635325,0.0003999584,0.002199894,0.0002467587,0.000498634],"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":[0.00002142809,0.00005252233,0.0000931694,0.0001219477,0.00008749055,0.00003066758,0.0003701131,0.9032653,0.0002334594,0.07659321,0.0007694544,0.01836126],"study_design_scores_gemma":[0.0004471128,0.0008281612,0.0003014703,0.00009264033,0.00001925071,0.000009026143,0.00001423876,0.9926782,0.0008480276,0.002087388,0.002088583,0.0005859084],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.000296343,0.00001012037,0.9636696,0.01913008,0.0002053947,0.0006452226,0.000008407073,0.001086658,0.01494819],"genre_scores_gemma":[0.23592,0.00004550955,0.7498739,0.00892464,0.001075421,0.00006947893,0.0002173244,0.00008381435,0.003789908],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.2356237,"threshold_uncertainty_score":0.9997829,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03920345239348902,"score_gpt":0.3165527323400888,"score_spread":0.2773492799465999,"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."}}