{"id":"W4399552717","doi":"10.1613/jair.1.15155","title":"Mitigating Value Hallucination in Dyna-Style Planning via Multistep Predecessor Models","year":2024,"lang":"en","type":"article","venue":"Journal of Artificial Intelligence Research","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Hallucinating; Computer science; Reinforcement learning; Bootstrapping (finance); Artificial intelligence; Value (mathematics); Function (biology); Successor cardinal; Sample (material); Bellman equation; Machine learning; Algorithm; Mathematical optimization; Econometrics; 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":["scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.01003265,0.0002564043,0.0004113842,0.0020666,0.000356219,0.001389593,0.001987313,0.0002043488,0.0000524834],"category_scores_gemma":[0.001417959,0.0002395502,0.0001873526,0.003315469,0.0002580677,0.003698358,0.0004160674,0.001783804,0.0001961936],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006673711,"about_ca_system_score_gemma":0.0006213395,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004407792,"about_ca_topic_score_gemma":0.0001579903,"domain_scores_codex":[0.9937815,0.0007647607,0.001734376,0.0006438794,0.00203045,0.001045046],"domain_scores_gemma":[0.9956499,0.001995911,0.0002729622,0.0005203468,0.001264336,0.0002965523],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00006659534,0.000244828,0.00005036403,0.0001293458,0.0000379905,0.001188686,0.0136209,0.2711004,0.01812792,0.1926791,0.0001103615,0.5026435],"study_design_scores_gemma":[0.00001874544,0.0002769712,0.00001824968,0.0006513834,0.000004067006,0.0001005594,0.002183583,0.7312263,0.07439232,0.1907828,0.0001723427,0.000172664],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05035209,0.001944141,0.9424772,0.00189509,0.0009667,0.0003629694,0.000001888512,0.00008402641,0.001915903],"genre_scores_gemma":[0.9563107,0.0001385211,0.04286009,0.00003842357,0.0004738994,0.00002163525,0.000001071776,0.00003378566,0.0001218312],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9059587,"threshold_uncertainty_score":0.9996471,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1987057900527737,"score_gpt":0.4453474910418657,"score_spread":0.246641700989092,"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."}}