{"id":"W6968549152","doi":"10.5281/zenodo.15285505","title":"Interpretable Concept Bottlenecks to Align Reinforcement Learning Agents","year":2024,"lang":"en","type":"article","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Reinforcement learning; Control (management); Key (lock); Process (computing); Action (physics); Matching (statistics)","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":[0.003343873,0.0009188123,0.00099013,0.001022229,0.0005338083,0.001447505,0.001822921,0.001605114,0.004919884],"category_scores_gemma":[0.01990644,0.0006826688,0.0005742052,0.000603441,0.001095824,0.002839651,0.002565082,0.00264514,0.0005537529],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001342817,"about_ca_system_score_gemma":0.001580379,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002422348,"about_ca_topic_score_gemma":0.0023374,"domain_scores_codex":[0.9988459,0.00048887,0.00008448668,0.0002563921,0.0002128632,0.0001113704],"domain_scores_gemma":[0.9933722,0.004199487,0.0005992448,0.0006554024,0.0008412588,0.0003324894],"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":[0.0007484051,0.0002780092,0.002342308,0.0001487901,0.0001172156,0.0001604822,0.0004706086,0.7281916,0.005660764,0.08454253,0.003042714,0.1742966],"study_design_scores_gemma":[0.00003301796,0.00003899635,0.00009728602,0.000008319288,0.00001116305,0.000007865973,0.00001531456,0.9725943,0.001010415,0.02578828,0.0003887428,0.000006303242],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03980665,0.0001853134,0.9560121,0.0003704671,0.0001221283,0.00008680439,0.0001016116,0.001400371,0.001914427],"genre_scores_gemma":[0.784709,0.00008249128,0.2125163,0.0001491818,0.00005411825,0.0002238301,0.0002499562,0.0002899562,0.001725017],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004919884,"threshold_uncertainty_score":0.01768434,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04140303531784949,"score_gpt":0.2817263463880848,"score_spread":0.2403233110702353,"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."}}