{"id":"W3139073295","doi":"10.48550/arxiv.2103.07945","title":"Learning One Representation to Optimize All Rewards","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Computer science; Representation (politics); Markov decision process; Artificial intelligence; Reinforcement learning; A priori and a posteriori; Range (aeronautics); Machine learning; Markov process; 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":[0.0008321302,0.001172396,0.001130474,0.0005010923,0.0003308882,0.001236038,0.001510304,0.001577221,0.004178167],"category_scores_gemma":[0.003796875,0.0004661395,0.0006622699,0.0005159405,0.0009843919,0.002280301,0.001581243,0.002239999,0.001047397],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001069974,"about_ca_system_score_gemma":0.001721202,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002102878,"about_ca_topic_score_gemma":0.002484579,"domain_scores_codex":[0.9994535,0.0001590507,0.00002504089,0.0001746269,0.00009655444,0.00009132223],"domain_scores_gemma":[0.9991881,0.0003584299,0.00008894473,0.0001827664,0.0001078767,0.00007381566],"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":[0.0001198217,0.0001123104,0.000606863,0.00009450939,0.00004573166,0.00006606287,0.00006715308,0.8520676,0.001987765,0.06094719,0.003031387,0.08085354],"study_design_scores_gemma":[0.00001342529,0.00003631898,0.00004976218,0.0000099106,0.000007680539,0.00001547573,0.000005037784,0.9704008,0.0006638508,0.02815738,0.0006328491,0.00000758119],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01263787,0.0001592806,0.983356,0.0004212517,0.0000471606,0.0000346318,0.0001552448,0.000668063,0.002520533],"genre_scores_gemma":[0.6806924,0.0004005587,0.3094014,0.0002978426,0.000105581,0.0003220692,0.000548801,0.0003175199,0.007913791],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004178167,"threshold_uncertainty_score":0.01397735,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1225321745249543,"score_gpt":0.2230164994008293,"score_spread":0.1004843248758749,"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."}}