{"id":"W4393883838","doi":"10.5281/zenodo.7901310","title":"APR-Models-Performance","year":2023,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Software System Performance and Reliability","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"","keywords":"Computer science","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","sts","scholarly_communication","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001329342,0.000305713,0.0003201877,0.0004500173,0.002409164,0.001356373,0.005155856,0.0002522967,0.0006979044],"category_scores_gemma":[0.0003071125,0.0002923342,0.0001065468,0.001270424,0.0001513421,0.0008399152,0.004253412,0.0006623042,0.1009778],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002293936,"about_ca_system_score_gemma":0.00001515719,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006552675,"about_ca_topic_score_gemma":5.135897e-7,"domain_scores_codex":[0.996937,0.0002932725,0.0004620065,0.0008934792,0.0007966155,0.0006175957],"domain_scores_gemma":[0.9968974,0.00004502799,0.0002212463,0.002011272,0.0006062463,0.0002188267],"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":[0.000008752402,0.00006005595,2.681091e-7,0.0003345516,0.00003038551,0.00001770621,0.0001148072,0.0001864003,0.000004803708,0.0001714934,0.9745579,0.02451292],"study_design_scores_gemma":[0.000234622,0.000183255,0.00005954232,0.0001157319,0.00001141359,0.0001180949,0.00001846946,0.004095185,0.00001427815,0.0001587361,0.9946529,0.0003377171],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0002221255,0.0001339758,0.0126496,0.0005373494,0.001710605,0.0009798282,0.9703266,0.004877945,0.008561988],"genre_scores_gemma":[0.0007335737,0.0007975537,0.0002994884,0.0001596278,0.0003614953,1.728806e-7,0.9958284,0.001026446,0.0007931729],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.1002799,"threshold_uncertainty_score":0.9999529,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04296245922289825,"score_gpt":0.2462466052606512,"score_spread":0.203284146037753,"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."}}