{"id":"W4382239271","doi":"10.1609/aaai.v37i6.25864","title":"The Sufficiency of Off-Policyness and Soft Clipping: PPO Is Still Insufficient according to an Off-Policy Measure","year":2023,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Optimization and Search Problems","field":"Computer Science","cited_by":15,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"Natural Sciences and Engineering Research Council of Canada; National Natural Science Foundation of China; Alberta Machine Intelligence Institute","keywords":"Clipping (morphology); Computer science; Space (punctuation); Metric (unit); Measure (data warehouse); Mathematical optimization; Sigmoid function; Code (set theory); Function (biology); Algorithm; Mathematics; Data mining; Artificial intelligence; Economics; Operations management; Programming language","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001740064,0.0002545928,0.0003105846,0.0004210253,0.0005111999,0.0005180555,0.002789411,0.00009528201,0.00001076349],"category_scores_gemma":[0.001687319,0.0001715523,0.00009512713,0.003101113,0.0004558829,0.000480804,0.00100382,0.0002918758,0.00004594006],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005163161,"about_ca_system_score_gemma":0.0003447621,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001100948,"about_ca_topic_score_gemma":0.00002141314,"domain_scores_codex":[0.9970674,0.00005154477,0.0007275185,0.0006237995,0.0009182186,0.0006115274],"domain_scores_gemma":[0.9972744,0.0002268883,0.000392239,0.0005148717,0.00135828,0.0002333136],"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":[0.00004535179,0.0001474676,0.0002918963,0.00004360942,0.00001208924,2.478488e-7,0.01143977,0.001017315,0.0138656,0.6697721,0.0001419703,0.3032226],"study_design_scores_gemma":[0.000085464,0.0008502186,0.000906812,0.0004140828,0.00001361433,0.000005891722,0.008536388,0.8101697,0.1290584,0.04783585,0.001627021,0.0004965565],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9184599,0.0001183891,0.03621808,0.0342982,0.0007301562,0.001824516,0.00002776178,0.0003908153,0.007932212],"genre_scores_gemma":[0.9982027,0.0002609326,0.0006902185,0.0004793606,0.00006445441,0.00003384925,4.527115e-7,0.00001714468,0.0002508466],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8091524,"threshold_uncertainty_score":0.6995698,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1051787284617897,"score_gpt":0.3426599136300764,"score_spread":0.2374811851682866,"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."}}