{"id":"W4393623462","doi":"10.5281/zenodo.8277054","title":"GEO-Bench: m-pv4ger-seg","year":2023,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"3D Modeling in Geospatial Applications","field":"Engineering","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","sts","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.0004178121,0.000311283,0.0002632965,0.0004048261,0.001304895,0.0006955346,0.001837179,0.0002586664,0.006464174],"category_scores_gemma":[0.0003425528,0.0003719523,0.00008840524,0.0007711484,0.0001059743,0.0001296413,0.001206854,0.0007623561,0.1708108],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002359692,"about_ca_system_score_gemma":0.000004714039,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001092916,"about_ca_topic_score_gemma":0.000003976873,"domain_scores_codex":[0.997954,0.0001028576,0.0004030265,0.0005240484,0.0004799547,0.0005360998],"domain_scores_gemma":[0.9981403,0.00003958197,0.00008512767,0.001154299,0.0003629653,0.0002176789],"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":[0.000003412872,0.00003546119,1.550427e-8,0.0001839382,0.00005927725,0.00001023339,0.00006090495,0.004733281,0.0001592469,0.0001091874,0.9876288,0.007016287],"study_design_scores_gemma":[0.0001544564,0.00004949909,0.00001062965,0.00005046591,0.00003700562,0.00003007598,0.000024911,0.005246464,0.00002705786,0.0001514624,0.9938578,0.0003601657],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.00003711717,0.00009144657,0.005106924,0.0002008405,0.0004527853,0.0005776412,0.9813501,0.004105615,0.008077528],"genre_scores_gemma":[0.000201666,0.0006314134,0.0002538258,0.00006533374,0.0004303614,3.319887e-7,0.9948969,0.003057495,0.0004627114],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.1643466,"threshold_uncertainty_score":0.9999953,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0359405857572664,"score_gpt":0.2402930168894324,"score_spread":0.204352431132166,"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."}}