{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01070945,0.002419853,0.001364495,0.005396304,0.0006914041,0.002688548,0.001423545,0.001807489,0.001343389],"category_scores_gemma":[0.1036208,0.0004988708,0.0007624247,0.003173936,0.0008647615,0.005771407,0.00122584,0.003141377,0.0009704262],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000669413,"about_ca_system_score_gemma":0.002065883,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008366936,"about_ca_topic_score_gemma":0.01174131,"domain_scores_codex":[0.992199,0.002971416,0.0007984788,0.001544845,0.001970602,0.0005156217],"domain_scores_gemma":[0.7917214,0.1678558,0.01785677,0.01194819,0.008146539,0.002471264],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0003408098,0.0003453481,0.6531615,0.0004363346,0.0002423946,0.0002408012,0.0003262199,0.1275582,0.003259852,0.001388613,0.003523558,0.2091764],"study_design_scores_gemma":[0.00003050356,0.0006394328,0.1480311,0.000334037,0.0002020529,0.0009740175,0.0004851307,0.8289295,0.006286752,0.010435,0.003525991,0.0001265072],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5304784,0.005900101,0.4435678,0.003718092,0.0003737929,0.0004070749,0.00339431,0.006636306,0.00552417],"genre_scores_gemma":[0.9318133,0.0006931559,0.06485224,0.0002258462,0.0001573428,0.00007104182,0.001574944,0.0001758064,0.0004363219],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01070945,"threshold_uncertainty_score":0.05663764,"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."}}