{"id":"W1992813097","doi":"10.1109/test.2014.7035339","title":"Clustering-based failure triage for RTL regression debugging","year":2014,"lang":"en","type":"article","venue":"","topic":"Formal Methods in Verification","field":"Computer Science","cited_by":17,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Debugging; Triage; Cluster analysis; Robustness (evolution); Data mining; Machine learning; Reliability engineering; Programming language; 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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001795243,0.001349409,0.001213421,0.004605146,0.00105811,0.001185693,0.00226385,0.001163492,0.002575056],"category_scores_gemma":[0.009319372,0.0005378592,0.001256688,0.002139824,0.000857042,0.001781944,0.001633994,0.001405449,0.001285509],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001311618,"about_ca_system_score_gemma":0.00175618,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005006563,"about_ca_topic_score_gemma":0.008926924,"domain_scores_codex":[0.997656,0.0004198721,0.0001875562,0.0005507147,0.0009759481,0.0002098533],"domain_scores_gemma":[0.9942464,0.001727748,0.001023381,0.001660345,0.001177682,0.0001645429],"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":[0.0005249583,0.0003334966,0.01795218,0.0004102516,0.0002059309,0.000438885,0.0009155815,0.3183067,0.02872944,0.01870378,0.01123155,0.6022473],"study_design_scores_gemma":[0.0000163804,0.00008622844,0.001702238,0.00003194825,0.00002846519,0.000192477,0.0001007457,0.9660307,0.01841857,0.009918545,0.003441131,0.000032518],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02239731,0.0001590966,0.9658521,0.0001154128,0.00002497718,0.0001488854,0.0003740166,0.01021373,0.0007145454],"genre_scores_gemma":[0.3719326,0.0001320215,0.623518,0.00008251152,0.00002643013,0.0002180462,0.001835869,0.0007530618,0.001501609],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005006563,"threshold_uncertainty_score":0.00995481,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0299304796440166,"score_gpt":0.3068266888396271,"score_spread":0.2768962091956105,"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."}}