{"id":"W1999095902","doi":"10.1109/cig.2013.6633642","title":"Stacked calibration of off-policy policy evaluation for video game matchmaking","year":2013,"lang":"en","type":"article","venue":"","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Ubisoft (Canada); Université de Montréal","funders":"","keywords":"Computer science; Function (biology); Calibration; Video game; Simple (philosophy); Statistics; Multimedia; 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004615623,0.0001074496,0.0001384426,0.0002752856,0.00005585636,0.000182366,0.0004507523,0.00004881064,0.0001140101],"category_scores_gemma":[0.0005139281,0.00009732493,0.00005879716,0.0004782983,0.00002945645,0.001057261,0.0001303161,0.00005189363,0.00004337052],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001298848,"about_ca_system_score_gemma":0.0003901295,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004352527,"about_ca_topic_score_gemma":0.000003989676,"domain_scores_codex":[0.9986354,0.00007110781,0.0003712065,0.0002131619,0.0004595482,0.0002495503],"domain_scores_gemma":[0.9987107,0.000146781,0.00024812,0.0004529857,0.0003846041,0.00005680421],"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":[0.000003968408,0.00002097543,0.0002189858,0.00004991825,0.00002235206,9.951317e-8,0.001036535,0.562798,0.002869276,0.3837174,0.002084884,0.04717755],"study_design_scores_gemma":[0.0003891884,0.00009493505,0.0008839879,0.00001545391,0.000005630007,0.000001299509,0.00003073956,0.9860126,0.002980471,0.008685154,0.0007966855,0.0001038985],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008540989,0.00001016666,0.9836513,0.002349518,0.0001033292,0.0008571207,6.268738e-7,0.0001063925,0.004380543],"genre_scores_gemma":[0.9387755,0.000003458101,0.0587288,0.0007894255,0.0001715142,0.00009808203,0.000009739681,0.00001181561,0.001411705],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9302345,"threshold_uncertainty_score":0.3968796,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03258298194827107,"score_gpt":0.321816013989087,"score_spread":0.289233032040816,"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."}}