{"id":"W4243478513","doi":"10.1007/978-1-4939-7131-2_100019","title":"Algorithm Evaluation","year":2018,"lang":"en","type":"book-chapter","venue":"","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Calgary","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":["insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.000588206,0.0001426169,0.0001123581,0.0001079257,0.0000690185,0.0001297713,0.0006009015,0.0001508654,0.003275176],"category_scores_gemma":[0.00003020153,0.0001245985,0.00004755493,0.00002422259,0.00002979999,0.0001864105,0.0001584384,0.0001660038,0.004557589],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005206081,"about_ca_system_score_gemma":0.0001036375,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000007805897,"about_ca_topic_score_gemma":0.000004768897,"domain_scores_codex":[0.9987808,0.00002561207,0.0001669871,0.0004304011,0.0004977122,0.00009852417],"domain_scores_gemma":[0.9986671,0.00003066323,0.0001410411,0.0008856273,0.0002269778,0.00004863326],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[1.548519e-7,0.000002048836,2.915896e-7,0.000001759451,0.000005997318,4.058528e-7,0.00002139279,5.101361e-7,0.000001057002,0.2983621,0.01883305,0.6827712],"study_design_scores_gemma":[0.00006455889,0.00003167865,0.00002827703,0.00001488264,0.00001209221,0.00000469657,5.879633e-7,0.200586,0.000005585639,0.09733462,0.7017739,0.0001431668],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"other","genre_gemma":"other","genre_scores_codex":[7.361322e-8,0.00005714995,0.4631139,0.0003259236,0.0002261823,0.00008020678,0.000002627582,0.0001297746,0.5360641],"genre_scores_gemma":[0.00002566803,0.00002649836,0.1578851,0.0002143636,0.0003399587,0.000006753929,0.0001915483,0.00001316633,0.841297],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.6829408,"threshold_uncertainty_score":0.997636,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04064680106131379,"score_gpt":0.2923500542145264,"score_spread":0.2517032531532126,"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."}}