{"id":"W7016997676","doi":"","title":"Aligning Language Models Using Multi-Objective Deep Reinforcement Learning","year":2023,"lang":"en","type":"other","venue":"Brock University Digital Repository (Brock University)","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Brock University","funders":"","keywords":"Reinforcement learning; Helpfulness; Task (project management); Perspective (graphical); Deep learning; Natural language","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":[0.001366728,0.001159372,0.0008944418,0.0005438959,0.0003573544,0.0009162294,0.001397071,0.001129812,0.002826873],"category_scores_gemma":[0.005143748,0.0005966171,0.0006594562,0.0004539882,0.0007120439,0.00147853,0.001385851,0.002043594,0.000876755],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001161513,"about_ca_system_score_gemma":0.001324682,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006040959,"about_ca_topic_score_gemma":0.0087326,"domain_scores_codex":[0.9992996,0.0002740856,0.00003372625,0.0001895543,0.000126992,0.0000759658],"domain_scores_gemma":[0.9984925,0.0009238147,0.0001614547,0.0001262993,0.0002145675,0.00008134245],"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":[0.00006669135,0.000111291,0.0007866027,0.00005350778,0.00005126792,0.00008011102,0.00006533304,0.9113612,0.002064697,0.004243845,0.001761784,0.07935365],"study_design_scores_gemma":[0.000005059764,0.00001197931,0.00002679207,0.000002525168,0.000002602271,0.000004115317,0.000002866361,0.9976541,0.0003112317,0.001838017,0.0001384787,0.000002202632],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.04301013,0.0004247031,0.9493604,0.0004936975,0.00008270049,0.0001147376,0.0001630569,0.002750607,0.003600045],"genre_scores_gemma":[0.7702474,0.0002111429,0.2232369,0.0004125149,0.00005285846,0.0002476764,0.0005751941,0.0003505992,0.004665679],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.006040959,"threshold_uncertainty_score":0.01201159,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01784163513093853,"score_gpt":0.2126744729845214,"score_spread":0.1948328378535829,"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."}}