{"id":"W4403536618","doi":"10.1145/3691620.3695261","title":"The Importance of Accounting for Execution Failures when Predicting Test Flakiness","year":2024,"lang":"en","type":"article","venue":"","topic":"Software Testing and Debugging Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Ubisoft (Canada)","funders":"","keywords":"Computer science; Test (biology); Reliability engineering; Accounting; Engineering; Business","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":[0.0007469296,0.00007923297,0.00007931831,0.00004855438,0.0001824126,0.0003043489,0.0005241146,0.00003440865,0.000001229649],"category_scores_gemma":[0.001375222,0.00005079319,0.00005019043,0.000216781,0.00003647698,0.00031398,0.0001362056,0.0000685964,0.000001306531],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001801946,"about_ca_system_score_gemma":0.00005435793,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005763533,"about_ca_topic_score_gemma":0.00002160731,"domain_scores_codex":[0.9992111,0.00001022664,0.0002246555,0.0002267002,0.0001527463,0.000174528],"domain_scores_gemma":[0.9971858,0.002274568,0.00007457298,0.0003278947,0.0001197769,0.00001734194],"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":[0.00000740793,0.00005317672,0.2104221,0.0004860759,0.00004859408,0.00001013932,0.001775757,0.00004218108,0.00241354,0.2606347,0.2092156,0.3148907],"study_design_scores_gemma":[0.00009098023,0.0001043631,0.003112151,0.000397175,0.00001326429,0.0000297598,0.00003161789,0.6732932,0.007746405,0.2998498,0.01512664,0.0002046148],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.004659886,0.0005061749,0.9829404,0.001864407,0.0003160711,0.0001931823,0.000002474231,0.008636754,0.0008806838],"genre_scores_gemma":[0.7729821,0.00000402127,0.2265791,0.00007120825,0.0001085278,0.00005246472,0.000001049476,0.000008688862,0.0001928975],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7683222,"threshold_uncertainty_score":0.2934843,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01749802715021794,"score_gpt":0.2739698035962184,"score_spread":0.2564717764460005,"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."}}