{"id":"W4393677065","doi":"10.5281/zenodo.7455766","title":"ATM: Black-box Test Case Minimization based on Test Code Similarity and Evolutionary Search – Replication Package","year":2022,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Software Testing and Debugging Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"","keywords":"Replication (statistics); Computer science; Code (set theory); Test (biology); Similarity (geometry); Programming language; Artificial intelligence; Biology; Mathematics; Statistics","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":[0.005472473,0.002160243,0.001109296,0.002518216,0.0005542057,0.001981944,0.004015624,0.00136594,0.1050683],"category_scores_gemma":[0.0346564,0.001436865,0.001959731,0.001514829,0.0006071546,0.002493591,0.002391607,0.001960223,0.04902298],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008705379,"about_ca_system_score_gemma":0.002289122,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003059322,"about_ca_topic_score_gemma":0.002021471,"domain_scores_codex":[0.9951583,0.001127691,0.0003959449,0.000689825,0.002386587,0.0002416329],"domain_scores_gemma":[0.9857567,0.00561246,0.0007779684,0.004168666,0.003388716,0.0002955422],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001190834,0.0002514697,0.002929372,0.0006865602,0.0002152149,0.0003125892,0.0001404476,0.06876968,0.01116406,0.02244546,0.4984357,0.3934586],"study_design_scores_gemma":[0.0007463464,0.0005161887,0.002517176,0.0001954549,0.0001132165,0.0006000848,0.00005624272,0.7520556,0.0278401,0.04179626,0.1733913,0.0001719208],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"methods","genre_gemma":"dataset","genre_scores_codex":[0.002950781,0.0001304771,0.742345,0.0004371174,0.0002443228,0.0007443731,0.008870228,0.2331686,0.01110905],"genre_scores_gemma":[0.07439568,0.0002073538,0.7946454,0.0006376796,0.0002801591,0.002860753,0.04116856,0.06145154,0.02435281],"genre_candidate":"dataset","genre_consensus":null,"teacher_disagreement_score":0.1050683,"threshold_uncertainty_score":0.3514884,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04402094876027669,"score_gpt":0.2760535441570954,"score_spread":0.2320325953968187,"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."}}