{"id":"W4402442387","doi":"10.1145/3650212.3680359","title":"ThinkRepair: Self-Directed Automated Program Repair","year":2024,"lang":"en","type":"article","venue":"","topic":"Software Testing and Debugging Techniques","field":"Computer Science","cited_by":40,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"","keywords":"Computer science","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[{"model":"gemma","categories":[],"domain":null,"study_design":"simulation_or_modeling","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"},{"model":"gpt","categories":[],"domain":null,"study_design":"not_applicable","genre":"software","about_ca_system":false,"about_ca_topic":false,"confidence":"medium","status":"direct model label, unvalidated"}],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002664501,0.004052534,0.001415622,0.002626467,0.0006667817,0.001434595,0.007179865,0.002692403,0.005716774],"category_scores_gemma":[0.01334521,0.001435561,0.003057632,0.001435423,0.001236486,0.0039383,0.003870714,0.003478885,0.006549443],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001031116,"about_ca_system_score_gemma":0.002329821,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007994171,"about_ca_topic_score_gemma":0.01677577,"domain_scores_codex":[0.9958979,0.001033349,0.0002728017,0.001587634,0.001006599,0.0002015465],"domain_scores_gemma":[0.9917514,0.003610448,0.0006058133,0.002873881,0.0008642214,0.0002943022],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001399698,0.0009435675,0.0093181,0.002551502,0.0004197911,0.0006845357,0.0009475555,0.03150733,0.02298512,0.004130649,0.325198,0.599914],"study_design_scores_gemma":[0.0008838839,0.0007568881,0.003775141,0.0001891802,0.0002470671,0.0009525631,0.0004008047,0.8319394,0.03734174,0.01549997,0.1078175,0.0001958772],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"software","genre_gemma":"methods","genre_scores_codex":[0.01515827,0.001229236,0.1860405,0.0005307836,0.0002375308,0.0003909406,0.006199448,0.7884814,0.001732008],"genre_scores_gemma":[0.1353701,0.0007297134,0.7703008,0.001995988,0.0001476298,0.0007974615,0.06436715,0.01831222,0.007978952],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.007994171,"threshold_uncertainty_score":0.01912451,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01264115361866649,"score_gpt":0.2930093712820749,"score_spread":0.2803682176634085,"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."}}