{"id":"W4378771708","doi":"10.48550/arxiv.2305.18246","title":"Provable and Practical: Efficient Exploration in Reinforcement Learning via Langevin Monte Carlo","year":2023,"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":"","funders":"Alberta Machine Intelligence Institute; Canadian Institute for Advanced Research","keywords":"Reinforcement learning; Markov chain Monte Carlo; Computer science; Monte Carlo method; Markov decision process; Posterior probability; Mathematical optimization; Scalability; Gaussian; Regret; Artificial intelligence; Algorithm; Markov process; Mathematics; Machine learning; Bayesian probability; Physics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"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":[0.0006440356,0.0002835907,0.0002995458,0.0004193039,0.0001479686,0.0002226558,0.0005155067,0.0002208459,0.000006618061],"category_scores_gemma":[0.0001899246,0.0003331853,0.00006374568,0.0006473116,0.00006325672,0.0005312535,0.00219391,0.0009348921,0.0001094394],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003429963,"about_ca_system_score_gemma":0.0001556058,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004246755,"about_ca_topic_score_gemma":0.00005869233,"domain_scores_codex":[0.9979554,0.000197365,0.0003104653,0.0009186255,0.0001976961,0.0004204375],"domain_scores_gemma":[0.9985144,0.0001761929,0.0003442084,0.0007211298,0.0001195557,0.0001245787],"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":[0.0000151669,0.00002085247,0.001615772,0.00009332656,0.00002830598,0.0002555032,0.0007222954,0.98861,0.00001697113,0.008428883,0.00008626273,0.0001066244],"study_design_scores_gemma":[0.0003754066,0.0001104054,0.0005765515,0.0001522694,0.0000250965,0.000003218875,0.0001941165,0.9972664,0.00003126231,0.0005563769,0.0003813023,0.0003276214],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04305793,0.00001875506,0.9543692,0.0003262725,0.0003662472,0.000568492,4.083645e-7,0.0003738509,0.0009187884],"genre_scores_gemma":[0.9910642,0.0002040119,0.002469538,0.00003874411,0.00003277259,0.000005705293,0.00001207198,0.00002649355,0.006146491],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9518997,"threshold_uncertainty_score":0.999912,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1289487721838383,"score_gpt":0.2268522974340941,"score_spread":0.09790352525025586,"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."}}