{"id":"W6969181150","doi":"10.5522/04/26531686.v1","title":"PitVis Challenge: Endoscopic Pituitary Surgery videos","year":2024,"lang":"en","type":"dataset","venue":"University College London","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Engineering and Physical Sciences Research Council; National Institute for Health and Care Research; Wellcome Trust","keywords":"Workflow; Annotation; Metadata; Scripting language; Endoscopic surgery; Multidisciplinary team; Baseline (sea)","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.0005683912,0.001063996,0.001418407,0.003125338,0.0005407727,0.00006326295,0.001505008,0.001178026,0.004261668],"category_scores_gemma":[0.0001734208,0.001310379,0.0008089172,0.001990591,0.0003331253,0.0004560247,0.001251889,0.001963278,0.09089126],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0019468,"about_ca_system_score_gemma":0.001193549,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003020123,"about_ca_topic_score_gemma":0.005327297,"domain_scores_codex":[0.9951335,0.0005655322,0.0005138171,0.001617174,0.001090968,0.00107901],"domain_scores_gemma":[0.9961603,0.0009240757,0.0003721055,0.001844004,0.0002262934,0.0004732407],"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":[0.0002470953,0.0003082073,0.0000244493,0.001105951,0.0005639114,0.0157725,0.0000416869,0.000003839396,0.000008329123,0.0002159715,0.9816043,0.0001038091],"study_design_scores_gemma":[0.000810405,0.0001180297,0.00004941802,0.0007616048,0.001057114,0.00004725223,0.0003669106,0.00002395138,0.00001495798,0.00005471495,0.9956857,0.001009944],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0005851403,0.002631836,0.000002369798,0.0005121717,0.002426818,0.0009392279,0.9903038,0.0007196023,0.00187903],"genre_scores_gemma":[0.0004566272,0.003033765,0.00003698273,0.00009150495,0.0006038757,0.000004337922,0.9836863,0.0001909193,0.01189574],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.08662958,"threshold_uncertainty_score":0.9989346,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01528757782809919,"score_gpt":0.2171137684283541,"score_spread":0.2018261906002549,"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."}}