{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0001040268,0.00007151906,0.00009143926,0.00005066422,0.00007234014,0.00003541113,0.00005486999,0.00004596645,0.00002293868],"category_scores_gemma":[0.0001163875,0.00004871999,0.00002215066,0.0001606,0.00006070013,0.00007853437,0.00001813697,0.00007748602,9.107661e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005688729,"about_ca_system_score_gemma":0.000009050943,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003215326,"about_ca_topic_score_gemma":0.0001684351,"domain_scores_codex":[0.9996372,0.00001193712,0.0001149941,0.00008213404,0.00005676849,0.0000969668],"domain_scores_gemma":[0.9995354,0.0002993417,0.00001103815,0.0001200377,0.00002104749,0.00001315089],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.000004204788,0.000006319277,0.001530519,0.0001411089,0.0000292302,3.562841e-7,0.000446779,0.01693953,0.0002150231,0.9748664,0.005517723,0.0003027887],"study_design_scores_gemma":[0.001807189,0.0002751983,0.004464123,0.001905588,0.0001930977,0.000005260753,0.05092803,0.2068958,0.04875289,0.6606558,0.02305537,0.001061612],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3532319,0.0007594527,0.5431973,0.01495978,0.0005750926,0.0008778151,0.00001005353,0.0004860529,0.08590255],"genre_scores_gemma":[0.9986329,0.0003317957,0.0008047963,0.00005759114,0.000007217109,0.00001978334,9.421563e-7,0.000006165436,0.0001387887],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.645401,"threshold_uncertainty_score":0.1986744,"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."}}