{"id":"W3167859469","doi":"10.5194/egusphere-egu21-3102","title":"Progress on a comprehensive earth system model evaluation framework","year":2021,"lang":"en","type":"article","venue":"","topic":"Hydrology and Watershed Management Studies","field":"Environmental Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Environment and Climate Change Canada; Saskatchewan Health; University of Saskatchewan","funders":"","keywords":"Computer science; Surprise; Benchmarking; Limit (mathematics); Process (computing); Predictability; Code (set theory); Work (physics); Operations research; Set (abstract data type); Mathematics; Engineering; Statistics; Programming language","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":[0.167362,0.002763108,0.003944588,0.009575829,0.002409477,0.01784297,0.01334258,0.003800912,0.008476858],"category_scores_gemma":[0.1560396,0.001359674,0.003533525,0.008588647,0.003873959,0.01888836,0.01043249,0.007427072,0.002392842],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.01311243,"about_ca_system_score_gemma":0.02593531,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.03179805,"about_ca_topic_score_gemma":0.02132934,"domain_scores_codex":[0.9083604,0.04638434,0.00666886,0.00304354,0.03303444,0.002508528],"domain_scores_gemma":[0.8756969,0.03665867,0.004696211,0.02319452,0.05721386,0.002539754],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001280081,0.0006958774,0.005536224,0.001451551,0.0004172432,0.0001289654,0.0006457291,0.3130412,0.001270712,0.453229,0.03763657,0.185819],"study_design_scores_gemma":[0.0001369385,0.0007292481,0.005353124,0.005456875,0.000251424,0.0001358238,0.001714775,0.6016201,0.005633186,0.2344726,0.1441916,0.0003044165],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0188252,0.004995549,0.8652477,0.02164154,0.0007495742,0.001796616,0.004279395,0.007656557,0.07480785],"genre_scores_gemma":[0.1721372,0.00475397,0.7974248,0.002505865,0.0004453133,0.002363067,0.01105908,0.003432037,0.005878714],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.167362,"threshold_uncertainty_score":0.8851056,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03130761308863122,"score_gpt":0.2763106151471577,"score_spread":0.2450030020585265,"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."}}