{"id":"W6973820966","doi":"10.57745/tvhvwl","title":"benchmark_v6.ipynb","year":2023,"lang":"en","type":"dataset","venue":"Recherche Data Gouv France","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Canadian Nautical Research Society","funders":"","keywords":"Workflow; Benchmark (surveying); Process (computing); Identification (biology); Work (physics)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.002231327,0.006610697,0.002331016,0.006249267,0.001331853,0.003613482,0.005164129,0.003326436,0.1001822],"category_scores_gemma":[0.009176189,0.001703073,0.003078697,0.00609158,0.0007812872,0.001977313,0.003092909,0.00310326,0.1295389],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002647725,"about_ca_system_score_gemma":0.003607139,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02439446,"about_ca_topic_score_gemma":0.03664388,"domain_scores_codex":[0.9979283,0.0003712702,0.0002041814,0.0006405847,0.0005384876,0.0003171548],"domain_scores_gemma":[0.9972038,0.0008499093,0.0001340623,0.0008615914,0.0006987248,0.0002518731],"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":[0.0001419614,0.00005438563,0.0004254195,0.0005748175,0.00005433208,0.00003297189,0.00002011852,0.0008129303,0.0004704223,0.000619569,0.9928974,0.003895601],"study_design_scores_gemma":[0.001155853,0.0001327196,0.003659939,0.0002902315,0.00008461116,0.0002296872,0.0001276235,0.006735598,0.004692133,0.005380654,0.9774053,0.0001056273],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0006765983,0.0001784341,0.001095304,0.0001715465,0.0001164468,0.0000984563,0.9780861,0.01678325,0.002793924],"genre_scores_gemma":[0.0007533437,0.00005182573,0.001860613,0.00007180263,0.00001185196,0.000194751,0.9947425,0.001317528,0.000995814],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.8998178,"threshold_uncertainty_score":0.3351429,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.5131220586473684,"score_gpt":0.4689175250193992,"score_spread":0.04420453362796922,"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."}}