{"id":"W7134485010","doi":"","title":"Metacognitive capabilities of LLMs: An exploration in mathematical problem solving","year":2024,"lang":"en","type":"article","venue":"HAL (Le Centre pour la Communication Scientifique Directe)","topic":"Innovative Teaching and Learning Methods","field":"Psychology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Metacognition; Comprehension; Perception; Identification (biology)","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":[0.0008992485,0.0002538839,0.000228261,0.0008299153,0.0002229731,0.002789934,0.0003862635,0.0002813468,0.003931977],"category_scores_gemma":[0.0139423,0.0001249647,0.0002072324,0.0007124155,0.000576573,0.001959757,0.0009304025,0.0007080535,0.0002682191],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006571881,"about_ca_system_score_gemma":0.0006172015,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001416778,"about_ca_topic_score_gemma":0.0008553458,"domain_scores_codex":[0.9995878,0.0001783144,0.00002720048,0.00007190763,0.00009098529,0.00004375254],"domain_scores_gemma":[0.9923706,0.005868645,0.0005990401,0.0005328575,0.0003904982,0.0002382796],"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":[0.002126135,0.001731031,0.2584631,0.001280776,0.0001800722,0.0009018702,0.05333448,0.007566931,0.03836926,0.06556376,0.001736766,0.5687458],"study_design_scores_gemma":[0.0001753167,0.001684117,0.7044263,0.0006260443,0.0002812351,0.001938928,0.02772659,0.05959233,0.0216873,0.1632798,0.0184384,0.000143605],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9731684,0.0006691438,0.007181187,0.0002712617,0.000009890171,0.000048232,0.0001065035,0.00005828669,0.01848714],"genre_scores_gemma":[0.9971982,0.0001357428,0.001642068,0.00002345987,0.000005043248,0.0000242514,0.00003162186,0.000008291978,0.0009314411],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.003931977,"threshold_uncertainty_score":0.01315373,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0507868683606436,"score_gpt":0.33862087558485,"score_spread":0.2878340072242064,"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."}}