{"id":"W4407694416","doi":"10.1067/j.cpradiol.2025.02.002","title":"Differentiating between GPT-generated and human-written feedback for radiology residents","year":2025,"lang":"en","type":"article","venue":"Current Problems in Diagnostic Radiology","topic":"Radiology practices and education","field":"Medicine","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University","funders":"Association Canadienne des Radiologistes","keywords":"Medicine; Radiology; Medical physics","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":true,"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":[0.05213586,0.0006746273,0.00053409,0.002377275,0.0007678717,0.002069411,0.0007915628,0.001023157,0.001969937],"category_scores_gemma":[0.2719802,0.0002635085,0.0005120758,0.0009694203,0.001418562,0.001568434,0.00253658,0.0007372156,0.001042434],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001968778,"about_ca_system_score_gemma":0.002980667,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004277626,"about_ca_topic_score_gemma":0.008088342,"domain_scores_codex":[0.934998,0.04090299,0.005142842,0.004153533,0.01372074,0.001081737],"domain_scores_gemma":[0.6900967,0.1958032,0.03844289,0.01160299,0.06002915,0.004025112],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.003974316,0.0005077578,0.4131089,0.003741425,0.0003465267,0.0006738764,0.09109458,0.003043838,0.03051318,0.0007179392,0.009147605,0.4431302],"study_design_scores_gemma":[0.0004391083,0.004683511,0.827679,0.003088843,0.0004237144,0.002177575,0.04307656,0.03541842,0.04951782,0.002581534,0.03037669,0.0005371337],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9655592,0.0005007169,0.02487312,0.0008341889,0.0001950401,0.00111134,0.0007241914,0.0007704839,0.005431724],"genre_scores_gemma":[0.9714624,0.0002284268,0.02531395,0.0004530639,0.00006950425,0.0007429549,0.000556517,0.0001348858,0.001038352],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.05213586,"threshold_uncertainty_score":0.2757241,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05103409684988507,"score_gpt":0.3699537446871576,"score_spread":0.3189196478372726,"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."}}