{"id":"W2466656563","doi":"","title":"Goodness of Fit of Skills Assessment Approaches: Insights from Patterns of Real vs. Synthetic Data Sets.","year":2015,"lang":"en","type":"article","venue":"PolyPublie (École Polytechnique de Montréal)","topic":"Intelligent Tutoring Systems and Adaptive Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"","keywords":"Goodness of fit; Synthetic data; Data set; Computer science; Set (abstract data type); Data modeling; Data mining; Signature (topology); Artificial intelligence; Machine learning; Mathematics; Database","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009745027,0.0002967953,0.0007064171,0.0003206509,0.00005855698,0.0000777652,0.002456068,0.0002036455,0.00000671085],"category_scores_gemma":[0.0001658309,0.0002697009,0.0001305732,0.0003953555,0.00007765539,0.0006512327,0.001184349,0.0003084531,0.00000278716],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002034105,"about_ca_system_score_gemma":0.0004145672,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.03261462,"about_ca_topic_score_gemma":0.0006061366,"domain_scores_codex":[0.9969398,0.0003251384,0.0009275803,0.0006611397,0.0007617459,0.0003845878],"domain_scores_gemma":[0.9958684,0.0002453034,0.0008714891,0.002512055,0.00028635,0.0002164073],"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":[0.0001298202,0.002774179,0.3208167,0.0006539405,0.0005190336,0.00007863138,0.01360366,0.01953101,0.01163542,0.5850354,0.0007711299,0.04445107],"study_design_scores_gemma":[0.001286918,0.0008990412,0.2132678,0.001539985,0.000123755,0.00004032925,0.002292278,0.6726059,0.1003067,0.004491168,0.002103028,0.001043109],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.25889,0.000313075,0.739104,0.0001511412,0.0002107434,0.0003967435,0.0001510347,0.0001299677,0.0006532069],"genre_scores_gemma":[0.9083036,0.00004786635,0.09116178,0.00004107136,0.00005743149,0.00004836587,0.00005043546,0.00003052483,0.0002589173],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.6530749,"threshold_uncertainty_score":0.9999755,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05883629423131383,"score_gpt":0.2767849305301021,"score_spread":0.2179486362987882,"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."}}