{"id":"W4377971347","doi":"10.1109/tai.2023.3279057","title":"Interpreting Tangled Program Graphs Under Partially Observable Dota 2 Invoker Tasks","year":2023,"lang":"en","type":"article","venue":"IEEE Transactions on Artificial Intelligence","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Computer science; Interpretability; Artificial intelligence; Graph; Context (archaeology); Machine learning; Task (project management); Theoretical computer science","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":[0.0008879534,0.0006710338,0.0003206499,0.0006684771,0.000421333,0.001586246,0.001019805,0.0008848854,0.003179056],"category_scores_gemma":[0.006252816,0.0004002334,0.0007991205,0.0002604088,0.001992469,0.002800152,0.002137968,0.001353609,0.0003111324],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001376548,"about_ca_system_score_gemma":0.00102544,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005632783,"about_ca_topic_score_gemma":0.007143462,"domain_scores_codex":[0.9991906,0.0002948894,0.00005764942,0.0001978246,0.0001646011,0.0000944065],"domain_scores_gemma":[0.9972862,0.001464026,0.0003338475,0.00053209,0.0002468778,0.0001369414],"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":[0.0001866561,0.0001253778,0.003395136,0.0001797301,0.00004451222,0.001008411,0.001745818,0.6727629,0.01047901,0.266059,0.001054221,0.04295916],"study_design_scores_gemma":[0.00002087235,0.00005182957,0.0002903024,0.00002509584,0.00001845546,0.00006734954,0.0001839262,0.7853457,0.004869274,0.2060694,0.003039346,0.00001840821],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1213929,0.00009171293,0.8686894,0.0003700047,0.00002916548,0.0001375384,0.0004036624,0.002562775,0.006322893],"genre_scores_gemma":[0.7806253,0.0001410344,0.2133932,0.0001441673,0.00001627257,0.0002338346,0.0006097201,0.0004623196,0.004374132],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005632783,"threshold_uncertainty_score":0.01120001,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08019492026766367,"score_gpt":0.3238346140799372,"score_spread":0.2436396938122736,"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."}}