{"id":"W4411272189","doi":"10.1109/icse-companion66252.2025.00075","title":"Revisiting SWE-Bench: On the Importance of Data Quality for LLM-Based Code Models","year":2025,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"York University","funders":"","keywords":"Computer science; Code (set theory); Quality (philosophy); Programming language; Software engineering; Physics","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":[0.03167545,0.001717825,0.001093204,0.003222199,0.001440293,0.006022414,0.004378398,0.00199154,0.001821866],"category_scores_gemma":[0.2220938,0.0008409318,0.001608276,0.003043341,0.002839996,0.008654105,0.004636491,0.006013649,0.001217813],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002163796,"about_ca_system_score_gemma":0.004283668,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01404296,"about_ca_topic_score_gemma":0.01952759,"domain_scores_codex":[0.9714271,0.01235256,0.002534165,0.004435458,0.008485151,0.0007656102],"domain_scores_gemma":[0.8034922,0.1321223,0.005840202,0.03891186,0.01761704,0.002016288],"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":[0.002424858,0.001707979,0.176505,0.005622926,0.001206886,0.001053633,0.005201026,0.1399554,0.02165789,0.03110524,0.1800352,0.4335239],"study_design_scores_gemma":[0.0004757225,0.00140568,0.04881861,0.002387098,0.0003381855,0.001411028,0.003469158,0.7686021,0.02498863,0.05235306,0.09542724,0.0003235531],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.654353,0.01504279,0.2231017,0.02740512,0.002401935,0.001007177,0.03225955,0.03182192,0.01260688],"genre_scores_gemma":[0.7510269,0.00238499,0.1820175,0.004751415,0.0004876802,0.0006045588,0.05164558,0.005002678,0.002078834],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03167545,"threshold_uncertainty_score":0.1675178,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1426113512617812,"score_gpt":0.4045948737246982,"score_spread":0.261983522462917,"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."}}