{"id":"W4394038155","doi":"10.5281/zenodo.8263732","title":"GEO-Bench: test-dataset","year":2023,"lang":"hu","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Computational Physics and Python Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Test (biology); Computer science; Environmental science; Biology; Ecology","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":[],"consensus_categories":[],"category_scores_codex":[0.001726601,0.005420387,0.001911604,0.004267926,0.00153451,0.00245433,0.004792398,0.003691978,0.05706901],"category_scores_gemma":[0.007049495,0.0009604535,0.002818009,0.004968259,0.001175642,0.00183052,0.002523314,0.002756866,0.09942736],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001808124,"about_ca_system_score_gemma":0.003419425,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02672731,"about_ca_topic_score_gemma":0.04983721,"domain_scores_codex":[0.997875,0.0003526849,0.0001949088,0.0006446786,0.0005644306,0.0003683529],"domain_scores_gemma":[0.996229,0.0008151926,0.0001902639,0.001316012,0.0009822379,0.0004673073],"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":[0.00007950014,0.00005300696,0.0003340339,0.0002730912,0.00003759524,0.00001916251,0.00001040933,0.0004659402,0.0002079023,0.0001994182,0.9962114,0.002108379],"study_design_scores_gemma":[0.001701401,0.0002265084,0.007759429,0.000335346,0.0001749784,0.000278664,0.0002183208,0.00718611,0.004515832,0.003852227,0.9735877,0.0001635088],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0009433423,0.0001276646,0.0003755651,0.0001963345,0.0001721387,0.00006423488,0.9921124,0.004642572,0.001365645],"genre_scores_gemma":[0.000715306,0.00003212337,0.0006775379,0.0000942379,0.00001665976,0.00009214113,0.9972759,0.0003242545,0.0007719111],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.05706901,"threshold_uncertainty_score":0.1909149,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05235524093384973,"score_gpt":0.2812710911375829,"score_spread":0.2289158502037331,"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."}}