{"id":"W1586383135","doi":"10.1109/latw.2015.7102521","title":"Exemplar-based failure triage for regression design debugging","year":2015,"lang":"en","type":"article","venue":"","topic":"VLSI and Analog Circuit Testing","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Debugging; Metric (unit); Triage; Flexibility (engineering); Process (computing); Data mining; Software bug; Reliability engineering; Cluster analysis; Machine learning; Programming language; Software; Engineering","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0007701307,0.0001085027,0.0001285344,0.00007916148,0.0001341748,0.0001491109,0.0004901703,0.00005029752,0.000005103018],"category_scores_gemma":[0.0003776846,0.00007995681,0.00005601178,0.0002229501,0.000009175756,0.0003173293,0.00005166155,0.00006259921,0.00002209913],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003182701,"about_ca_system_score_gemma":0.0001801928,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000008886891,"about_ca_topic_score_gemma":0.000003888517,"domain_scores_codex":[0.9990421,0.00006541323,0.0001551418,0.00028213,0.0001859817,0.0002692625],"domain_scores_gemma":[0.9990669,0.0002909239,0.00006557535,0.0003174549,0.0001095268,0.0001495498],"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":[0.00001010576,0.0001852601,0.001795603,0.00007026466,0.00002735489,0.00008597322,0.001176967,0.006009225,0.01859946,0.03498641,0.1957127,0.7413406],"study_design_scores_gemma":[0.002009425,0.0002344517,0.00004395998,0.0001106353,0.000009891987,0.00001781249,0.00007949297,0.9493801,0.02634651,0.01246052,0.008973702,0.0003334684],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0004638266,0.00006188761,0.9954221,0.001782718,0.0001519654,0.0002081786,3.96734e-7,0.0004551309,0.001453763],"genre_scores_gemma":[0.724048,1.527994e-7,0.2749965,0.0005342147,0.000065752,0.00002623541,0.000002084701,0.000009175127,0.0003178908],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9433709,"threshold_uncertainty_score":0.3260544,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1467957928144607,"score_gpt":0.3005540975466254,"score_spread":0.1537583047321647,"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."}}