{"id":"W3199199701","doi":"10.48550/arxiv.2106.00589","title":"Improving Long-Term Metrics in Recommendation Systems using Short-Horizon Reinforcement Learning","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Advanced Bandit Algorithms Research","field":"Decision Sciences","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Reinforcement learning; Term (time); Horizon; Computer science; Reinforcement; Recommender system; Artificial intelligence; Machine learning; 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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.003519417,0.0004017869,0.0007244952,0.002347994,0.0003120847,0.0008133372,0.001357997,0.0004446145,0.0002428217],"category_scores_gemma":[0.002369208,0.0004400468,0.0002479349,0.003802002,0.0001034323,0.001039103,0.002996253,0.001734577,0.00004048256],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002028683,"about_ca_system_score_gemma":0.0004875768,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007767502,"about_ca_topic_score_gemma":0.0002168636,"domain_scores_codex":[0.994953,0.0008642212,0.0009167229,0.001804889,0.0007780028,0.0006831056],"domain_scores_gemma":[0.9960737,0.001008957,0.0006791709,0.001071319,0.0009315503,0.0002353536],"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":[0.00004112984,0.00004202049,0.03875087,0.00007255573,0.00003777769,0.0004587152,0.00009618555,0.9431597,0.0001092655,0.0001033088,0.000006838579,0.01712163],"study_design_scores_gemma":[0.0004139188,0.00008787575,0.002158895,0.0001978235,0.00003912145,0.00000810547,0.001738544,0.9943711,0.0001675086,0.0002759998,0.00008345192,0.0004576787],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4241119,0.0001047237,0.5742497,0.000008132888,0.0008042459,0.0004041939,0.000004357499,0.00004582555,0.0002669305],"genre_scores_gemma":[0.9970151,0.000296152,0.0005599761,0.00000551457,0.0001426155,0.000002464325,0.0001135463,0.00003872776,0.001825945],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5736897,"threshold_uncertainty_score":0.9998052,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3178481876547598,"score_gpt":0.3384350574772259,"score_spread":0.02058686982246605,"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."}}