{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000874621,0.0002150782,0.0002523233,0.0005037065,0.0002378965,0.0001819303,0.0003380249,0.0001615948,0.000001797493],"category_scores_gemma":[0.002442514,0.0002262246,0.00005313505,0.0004642847,0.00003595054,0.0004917987,0.00003786025,0.0002850204,0.00000225375],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002967716,"about_ca_system_score_gemma":0.00005066157,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000183705,"about_ca_topic_score_gemma":0.00000684179,"domain_scores_codex":[0.9986416,0.00006825423,0.0001989007,0.0005049099,0.0001509207,0.000435481],"domain_scores_gemma":[0.990461,0.008895342,0.00002774684,0.0004246066,0.0000680262,0.0001232863],"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":[0.000008378643,0.00001760318,0.00008441402,0.0001582763,0.00006486614,0.0000140667,0.001764522,0.8968477,0.01406003,0.008473975,0.0001405063,0.07836566],"study_design_scores_gemma":[0.0007395998,0.0003874864,0.0006678272,0.00007296526,0.00002948917,0.0001462843,0.0002870038,0.9804311,0.007650285,0.006877863,0.002172746,0.000537355],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0482391,0.0001941159,0.9492288,0.0004457561,0.0003551288,0.0002097918,0.00001294354,0.001313805,5.350679e-7],"genre_scores_gemma":[0.3083429,0.00008097518,0.6912599,0.00002403658,0.00006529803,0.0001630077,0.000004294472,0.00002843664,0.00003113308],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.2601038,"threshold_uncertainty_score":0.9225172,"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."}}