{"id":"W4400680152","doi":"10.1109/saner60148.2024.00044","title":"CEDAR: Continuous Testing of Deep Learning Libraries","year":2024,"lang":"en","type":"article","venue":"","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"York University","funders":"National Science Foundation","keywords":"Computer science; Deep learning; Artificial intelligence","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":[],"consensus_categories":[],"category_scores_codex":[0.00005549654,0.00006071718,0.00008008262,0.00004555321,0.00006631399,0.000116529,0.0003336867,0.0000188669,0.00001522615],"category_scores_gemma":[0.00007006816,0.00005207322,0.0000245451,0.0006675262,0.00003780387,0.000478285,0.0001933103,0.0001159566,0.0000426014],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000005881263,"about_ca_system_score_gemma":0.00001802497,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000003449612,"about_ca_topic_score_gemma":9.533427e-7,"domain_scores_codex":[0.9994162,0.00001568706,0.0001389556,0.0002075216,0.00008917764,0.0001325077],"domain_scores_gemma":[0.9992129,0.0004957696,0.00003167742,0.0001948565,0.00003260703,0.00003218608],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[3.157327e-7,0.000005248098,0.0004877429,0.0000129244,0.000004141365,0.000005182798,0.0001549005,0.002607635,0.002879826,0.5308399,0.0001858974,0.4628163],"study_design_scores_gemma":[0.00005345931,0.00006898733,0.0005108789,0.00004308395,0.000004484973,0.00002836728,0.00004778599,0.8931899,0.01359821,0.04505294,0.04724105,0.0001608625],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.00234609,0.0006881232,0.9692597,0.0004398212,0.00008442954,0.00006446172,1.750988e-7,0.0007982507,0.0263189],"genre_scores_gemma":[0.6059408,0.000005233598,0.3931649,0.00005392733,0.00003641402,0.00000828483,4.870897e-7,0.000006141056,0.0007837451],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8905823,"threshold_uncertainty_score":0.2123484,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02153558194571753,"score_gpt":0.2542780030835139,"score_spread":0.2327424211377964,"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."}}