{"id":"W6969313300","doi":"10.5281/zenodo.8256421","title":"A Probabilistic Framework for Mutation Testing in Deep Neural Networks - Models archive Part 4","year":2022,"lang":"en","type":"article","venue":"PolyPublie (École Polytechnique de Montréal)","topic":"Generative Adversarial Networks and Image Synthesis","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"","keywords":"Probabilistic logic; Replication (statistics); Artificial neural network; Mutation; Software; Statistical model","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009488456,0.0002856672,0.0003283393,0.0002898565,0.0005910103,0.0002665716,0.00094477,0.000104964,0.000009568783],"category_scores_gemma":[0.0006442295,0.0003110849,0.0001399978,0.001091462,0.0000552015,0.0005951814,0.0005702731,0.0005882123,6.663132e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003333432,"about_ca_system_score_gemma":0.0001255468,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001376962,"about_ca_topic_score_gemma":0.000933136,"domain_scores_codex":[0.9973122,0.0003956382,0.0005055626,0.0006748884,0.0003123964,0.0007992404],"domain_scores_gemma":[0.9976916,0.001107962,0.0002560901,0.0006757012,0.0001030095,0.0001656486],"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":[0.00003878273,0.0001102621,0.0004066783,0.000009445205,0.00001271693,0.00003214068,0.0003945674,0.8734085,0.00006361841,0.08379374,0.0001992235,0.04153031],"study_design_scores_gemma":[0.0002472481,0.0002006239,0.0006783911,0.00001890654,0.00001240456,0.00004092175,0.0000672876,0.8224044,0.00005383105,0.1759003,0.0001025096,0.0002731183],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.002475398,0.0005144835,0.9932787,0.001601285,0.0003021436,0.001355924,0.0000250603,0.0003532605,0.00009370247],"genre_scores_gemma":[0.5494869,0.000004767026,0.4473828,0.0009613588,0.0001383348,0.001962701,0.00001873502,0.00002710948,0.00001727944],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.5470115,"threshold_uncertainty_score":0.9999341,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02181960590637508,"score_gpt":0.2323912388658238,"score_spread":0.2105716329594487,"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."}}