{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0008132873,0.0002844704,0.0008199236,0.0004352729,0.0002040845,0.00002822608,0.0001875571,0.000359368,0.00003024161],"category_scores_gemma":[0.005484419,0.0002612281,0.00008104119,0.000292643,0.0002898554,0.0001121617,0.00008474892,0.0005418413,0.000007753519],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001577942,"about_ca_system_score_gemma":0.0001567026,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009431333,"about_ca_topic_score_gemma":0.00003956151,"domain_scores_codex":[0.9974149,0.0003494337,0.0008681025,0.0006925036,0.00007608677,0.0005989253],"domain_scores_gemma":[0.9942954,0.00484645,0.0002844722,0.0003288723,0.0001161378,0.000128707],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00005077786,0.0001915054,0.9725155,0.0005653603,0.0001315629,0.000004528906,0.0002545947,0.00004991898,0.001062122,0.002757124,0.009531426,0.01288563],"study_design_scores_gemma":[0.003625616,0.0004359695,0.9827459,0.0005367258,0.0003368833,0.00008915567,0.00007437155,0.0002192397,0.0001262442,0.006854452,0.004728034,0.0002274125],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9806932,0.01085835,0.001939161,0.002893869,0.00163329,0.001607034,0.00001829882,0.00005852795,0.0002982892],"genre_scores_gemma":[0.9951221,0.002111462,0.0005308845,0.0002445271,0.0005878709,0.0005554123,0.0005769783,0.00002541346,0.000245383],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01442889,"threshold_uncertainty_score":0.999984,"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."}}