{"id":"W4390051453","doi":"10.1145/3638246","title":"Test Generation Strategies for Building Failure Models and Explaining Spurious Failures","year":2023,"lang":"en","type":"article","venue":"ACM Transactions on Software Engineering and Methodology","topic":"Software Engineering Research","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Spurious relationship; Computer science; Test (biology); Machine learning; Surrogate model; Test case; Reliability engineering; Artificial intelligence; 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":[0.002906146,0.001934566,0.0009377559,0.002180446,0.0004374303,0.001140663,0.002534576,0.001631492,0.001844851],"category_scores_gemma":[0.02162061,0.000809661,0.001571363,0.000871212,0.001203148,0.00170238,0.001402241,0.001803062,0.0005477121],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001240221,"about_ca_system_score_gemma":0.001693994,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004216533,"about_ca_topic_score_gemma":0.005732094,"domain_scores_codex":[0.9975466,0.001061394,0.0001651609,0.0004651248,0.0006240522,0.0001375762],"domain_scores_gemma":[0.9855763,0.01052239,0.0009777109,0.00151594,0.001222555,0.0001849861],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000135098,0.0001497031,0.006780132,0.0001474329,0.00009617845,0.0003534973,0.0001870282,0.88171,0.004394119,0.01047218,0.001723122,0.09385153],"study_design_scores_gemma":[0.00001023885,0.00002817299,0.0001812937,0.00001119901,0.00001183128,0.0000500783,0.00001149572,0.991126,0.001534839,0.006749371,0.0002795085,0.000005956097],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02616278,0.000162287,0.9700863,0.0002600458,0.00001726051,0.0001119109,0.0002482428,0.00219161,0.0007595545],"genre_scores_gemma":[0.5296723,0.0001812563,0.4668823,0.0002761988,0.00003051978,0.000389822,0.001356669,0.0003894409,0.0008215365],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004216533,"threshold_uncertainty_score":0.01536936,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1374619766859885,"score_gpt":0.3462646189624147,"score_spread":0.2088026422764261,"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."}}