{"id":"W4413360646","doi":"10.1109/icse-seip66354.2025.00023","title":"On the Diagnosis of Flaky Job Failures: Understanding and Prioritizing Failure Categories","year":2025,"lang":"en","type":"article","venue":"","topic":"Reliability and Maintenance Optimization","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Telus (Canada); École de Technologie Supérieure","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Reliability engineering; Engineering","routes":{"ca_aff":true,"ca_fund":true,"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":[0.002590957,0.001331269,0.0009578245,0.01102954,0.0010333,0.002247638,0.001356013,0.001115846,0.001365116],"category_scores_gemma":[0.01500579,0.000367509,0.0006410699,0.003756554,0.0008627475,0.002234242,0.00138494,0.001182624,0.0005795604],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001556226,"about_ca_system_score_gemma":0.002105261,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02633445,"about_ca_topic_score_gemma":0.03604463,"domain_scores_codex":[0.998014,0.0003659527,0.0001727948,0.0004248404,0.0006717235,0.0003507558],"domain_scores_gemma":[0.9854021,0.006742842,0.003608242,0.0006173908,0.00296385,0.0006654651],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.000716647,0.0007169382,0.5481548,0.0006494303,0.0002042163,0.0008327343,0.002298376,0.1424548,0.01399742,0.004722067,0.005102435,0.2801501],"study_design_scores_gemma":[0.00002707789,0.0004540114,0.2630885,0.0002243425,0.0000916395,0.0007124865,0.003816669,0.7090361,0.007877952,0.01049162,0.004054169,0.0001254525],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8169364,0.001643614,0.1734532,0.001155047,0.00008198443,0.000341282,0.001394709,0.0009210362,0.004072757],"genre_scores_gemma":[0.9474817,0.0002581569,0.050265,0.00008262102,0.00003154889,0.00005185753,0.0008735416,0.00006011552,0.0008955927],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02633445,"threshold_uncertainty_score":0.05236238,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01329707349776384,"score_gpt":0.208806746648095,"score_spread":0.1955096731503312,"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."}}