{"id":"W7102328828","doi":"","title":"Advantage Shaping as Surrogate Reward Maximization: Unifying Pass@K Policy Gradients","year":2025,"lang":"","type":"article","venue":"arXiv (Cornell University)","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Reinforcement learning; Perspective (graphical); Simple (philosophy); Surrogate model; Verifiable secret sharing; Optimization problem","routes":{"ca_aff":true,"ca_fund":true,"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":[0.003014327,0.00103152,0.001132093,0.0005207736,0.0004908541,0.001826453,0.001410244,0.001699361,0.004655837],"category_scores_gemma":[0.01165545,0.000518318,0.0006343987,0.0004176077,0.003156103,0.003064594,0.003434866,0.002848369,0.0009550569],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001552843,"about_ca_system_score_gemma":0.001935319,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0010755,"about_ca_topic_score_gemma":0.001174189,"domain_scores_codex":[0.9986131,0.0006035338,0.00005880651,0.0002287108,0.0003692718,0.0001264766],"domain_scores_gemma":[0.9969175,0.001905249,0.0002667171,0.0005026758,0.0002441349,0.0001637994],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001401353,0.00006545488,0.0005575706,0.0001326232,0.00004103232,0.0001017369,0.0002301382,0.3880112,0.003145802,0.5292142,0.002256542,0.07610342],"study_design_scores_gemma":[0.00001935608,0.00007035573,0.00008580933,0.00003923381,0.00001085345,0.00004068438,0.0000191205,0.8143511,0.001565922,0.1810177,0.002761336,0.00001858834],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.00957298,0.0001726503,0.9820895,0.0004824732,0.00004758685,0.00003876026,0.00002303429,0.000334721,0.007238176],"genre_scores_gemma":[0.7018767,0.0004520815,0.2875677,0.0005578368,0.0001106467,0.0002148508,0.00006476667,0.0005499048,0.008605552],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004655837,"threshold_uncertainty_score":0.01594144,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06663284614697988,"score_gpt":0.2247884175724866,"score_spread":0.1581555714255067,"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."}}