{"id":"W7123340585","doi":"10.1109/esem64174.2025.00044","title":"Mapping Code Smells and Refactorings Accurately: Insights from an Empirical Study","year":2025,"lang":"","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Code refactoring; Code smell; Code (set theory); Software; Software quality; Code review; Software maintenance; Source code","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":[0.03497767,0.0004454415,0.0004273413,0.005218081,0.0009032729,0.002659985,0.001429376,0.0008755791,0.001002452],"category_scores_gemma":[0.2825181,0.0005612869,0.0003958232,0.005548294,0.001558962,0.004858169,0.001900411,0.001669735,0.0003549681],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001950627,"about_ca_system_score_gemma":0.002473384,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008099049,"about_ca_topic_score_gemma":0.01372282,"domain_scores_codex":[0.9696552,0.01383229,0.003269515,0.00257774,0.009877778,0.0007875525],"domain_scores_gemma":[0.402124,0.4519846,0.06703277,0.01816058,0.0584048,0.002293257],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0001781805,0.0005883695,0.8703915,0.0008888877,0.000151948,0.0005757598,0.04158121,0.001233492,0.00197502,0.0007588496,0.001658094,0.0800187],"study_design_scores_gemma":[0.00003735633,0.0005727359,0.944504,0.0007225588,0.0001317642,0.0006557932,0.03015183,0.01023613,0.003059444,0.001309497,0.00852244,0.00009644837],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9925205,0.0006130221,0.004453055,0.0004215896,0.00001060673,0.00009909332,0.0002806998,0.00005326651,0.001548189],"genre_scores_gemma":[0.9944885,0.0003135852,0.004235826,0.00009549682,0.000009559172,0.00007629916,0.0003816989,0.00005016746,0.0003489133],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03497767,"threshold_uncertainty_score":0.1849818,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08044891314345476,"score_gpt":0.3775742519305696,"score_spread":0.2971253387871148,"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."}}