{"id":"W3163202066","doi":"10.1109/tse.2021.3082068","title":"An Empirical Study of Type-Related Defects in Python Projects","year":2021,"lang":"en","type":"article","venue":"IEEE Transactions on Software Engineering","topic":"Software Engineering Research","field":"Computer Science","cited_by":20,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; McGill University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Python (programming language); Computer science; Programming language; Artificial intelligence","routes":{"ca_aff":true,"ca_fund":true,"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":[0.03467667,0.0004719624,0.0005160925,0.005740139,0.001703361,0.002958979,0.00187463,0.001838991,0.002532884],"category_scores_gemma":[0.318782,0.0006614745,0.0005288449,0.005409111,0.002998511,0.005997955,0.003379532,0.00300269,0.0009051239],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002544001,"about_ca_system_score_gemma":0.002971521,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003851909,"about_ca_topic_score_gemma":0.004288912,"domain_scores_codex":[0.956293,0.01646803,0.005174926,0.004942703,0.01534587,0.001775563],"domain_scores_gemma":[0.4353019,0.3116716,0.157365,0.01945409,0.06721427,0.008993046],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002706387,0.0006142471,0.9561905,0.0002884651,0.0000630402,0.0003998847,0.0134012,0.0003461726,0.0003946144,0.0008605401,0.002330895,0.02483983],"study_design_scores_gemma":[0.0000606371,0.001019057,0.9527838,0.0004137062,0.00008172021,0.001075853,0.02897363,0.003953106,0.001297735,0.001294613,0.008951855,0.00009424749],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.994288,0.0002366121,0.001844022,0.0004015204,0.00002369885,0.000199494,0.0004966014,0.00009081378,0.002419278],"genre_scores_gemma":[0.995715,0.0001584418,0.001933995,0.0001466208,0.0000193122,0.0003808391,0.0006782156,0.0000660596,0.0009014664],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03467667,"threshold_uncertainty_score":0.18339,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02540647067592235,"score_gpt":0.2945407086543776,"score_spread":0.2691342379784553,"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."}}