{"id":"W4394028127","doi":"10.5281/zenodo.8008681","title":"GEO-Bench: m-so2sat","year":2023,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Geophysics and Gravity Measurements","field":"Earth and Planetary Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; Mila - Quebec Artificial Intelligence Institute","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0006289018,0.002155344,0.001176129,0.001858609,0.0006255741,0.001630042,0.002492291,0.001746504,0.04907858],"category_scores_gemma":[0.002305511,0.0006503952,0.001065272,0.004042074,0.0003924398,0.002013347,0.001636806,0.001708752,0.1170059],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009300723,"about_ca_system_score_gemma":0.001227523,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02081904,"about_ca_topic_score_gemma":0.02870445,"domain_scores_codex":[0.9993411,0.0001004646,0.00006833213,0.0002125184,0.0001607025,0.0001168542],"domain_scores_gemma":[0.9989478,0.0001201942,0.0000917116,0.0003609285,0.0003797289,0.00009968615],"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":[0.00004396362,0.00001729921,0.0008639669,0.000244205,0.00003444422,0.00002089024,0.00001894269,0.0005157353,0.0002752584,0.0004720283,0.9957609,0.001732473],"study_design_scores_gemma":[0.0001969528,0.00001316835,0.005735212,0.0001174664,0.00002561528,0.00004895109,0.0000968375,0.001291814,0.0009104537,0.001763199,0.9897527,0.00004756843],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0002179073,0.00003230078,0.0001716315,0.00006125257,0.00003346628,0.000008380209,0.9974032,0.001211076,0.0008607755],"genre_scores_gemma":[0.0006044642,0.00002087781,0.0003136931,0.0000360076,0.000009066059,0.00002423283,0.9983968,0.0001921709,0.0004026864],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.04907858,"threshold_uncertainty_score":0.1641842,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05788304442182647,"score_gpt":0.2357220886196294,"score_spread":0.1778390441978029,"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."}}