{"id":"W4389141459","doi":"10.1007/s10664-023-10389-6","title":"Silent bugs in deep learning frameworks: an empirical study of Keras and TensorFlow","year":2023,"lang":"en","type":"article","venue":"Empirical Software Engineering","topic":"Software Engineering Research","field":"Computer Science","cited_by":38,"is_retracted":false,"has_abstract":false,"ca_institutions":"Polytechnique Montréal","funders":"Fonds de recherche du Québec; Consortium de Recherche et d’innovation en Aérospatiale au Québec; Canadian Institute for Advanced Research","keywords":"Software bug; Computer science; Relevance (law); Debugging; Artificial intelligence; Empirical research; Deep learning; Machine learning; Software; Programming language","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0007966782,0.0003029027,0.0004548458,0.0007151951,0.00008178638,0.0001248895,0.0007526302,0.0002790673,0.00001044205],"category_scores_gemma":[0.004821883,0.0003057821,0.00006132034,0.002538047,0.00003277393,0.0004146755,0.0006485448,0.001401623,0.00002576711],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001163052,"about_ca_system_score_gemma":0.00003861107,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004442613,"about_ca_topic_score_gemma":0.0000107082,"domain_scores_codex":[0.9971774,0.0001312888,0.0004874872,0.0007404437,0.0007193772,0.0007440164],"domain_scores_gemma":[0.9962875,0.002639148,0.00004942103,0.0006112237,0.00009158458,0.000321154],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.000005796861,0.0001781613,0.7727258,0.00005062195,0.00001953247,0.0001267156,0.004600234,0.2171709,0.00004667343,0.000006688339,0.00005444929,0.005014414],"study_design_scores_gemma":[0.0004350948,0.0003959333,0.7361699,0.00005428299,0.000003791795,0.00001035494,0.0002004388,0.2621328,0.00003952887,0.00004142063,0.000247349,0.0002690229],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7940602,0.0001458286,0.2038906,0.0001179066,0.0001945214,0.0002545717,6.557017e-7,0.001334126,0.000001567591],"genre_scores_gemma":[0.9828213,0.00001931832,0.01688245,0.00003880433,0.00008467044,0.0000706112,0.000003735049,0.00005545393,0.00002359024],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1887611,"threshold_uncertainty_score":0.9999394,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02542377729338102,"score_gpt":0.3172849195316602,"score_spread":0.2918611422382792,"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."}}