{"id":"W4393895664","doi":"10.5281/zenodo.4028817","title":"AbdulAli/replication-kit-emse-2020-benchmark: Second Release","year":2020,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Radiomics and Machine Learning in Medical Imaging","field":"Medicine","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Replication (statistics); Benchmark (surveying); Computer science; Environmental science; Biology; Geology; Virology; Geodesy","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001881633,0.004544985,0.001591771,0.002793147,0.0009143475,0.002578753,0.004859732,0.003248501,0.05429349],"category_scores_gemma":[0.009366521,0.000945207,0.002264707,0.002797621,0.0006208902,0.002019921,0.001972919,0.002486327,0.08582332],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001836535,"about_ca_system_score_gemma":0.002350793,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02179091,"about_ca_topic_score_gemma":0.04164237,"domain_scores_codex":[0.9983799,0.0003263938,0.0001647228,0.000396389,0.0004981867,0.0002343729],"domain_scores_gemma":[0.9971922,0.0006348496,0.0001257056,0.0009622024,0.0007751078,0.0003101098],"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":[0.000328903,0.0001624858,0.0008031782,0.0005173922,0.00009954831,0.00005584351,0.00001346591,0.00416121,0.0004295312,0.0007411435,0.9834887,0.00919857],"study_design_scores_gemma":[0.001979467,0.0003295304,0.007669087,0.0003882642,0.0001394314,0.0003356617,0.0001363488,0.04611793,0.00629088,0.009465305,0.927024,0.0001240441],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.006383114,0.001101683,0.002923565,0.001019733,0.0008185593,0.0002357489,0.9330807,0.03984644,0.01459044],"genre_scores_gemma":[0.00518744,0.0001723459,0.004491196,0.0002473103,0.00004304795,0.0001832102,0.9842037,0.001693828,0.003777974],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.05429349,"threshold_uncertainty_score":0.1816298,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02168605006693855,"score_gpt":0.2792266635721622,"score_spread":0.2575406135052236,"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."}}