{"id":"W7161205419","doi":"","title":"Context is Key: A Benchmark for Forecasting with Essential Textual Information","year":2024,"lang":"","type":"article","venue":"PolyPublie (École Polytechnique de Montréal)","topic":"Forecasting Techniques and Applications","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Oak Ridge National Laboratory; Office of Science; Mitacs; U.S. Department of Energy","keywords":"Benchmark (surveying); Context (archaeology); Task (project management); Time series; Range (aeronautics); Variety (cybernetics); Series (stratigraphy)","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","sts","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.00380842,0.0009778737,0.0009567529,0.001693035,0.001347905,0.004800746,0.001876008,0.0007903164,0.0005410385],"category_scores_gemma":[0.001400833,0.0008245431,0.0007210734,0.003333264,0.0005547671,0.003217446,0.0005955172,0.0010346,0.0001888697],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007293081,"about_ca_system_score_gemma":0.001456937,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004823775,"about_ca_topic_score_gemma":0.002159553,"domain_scores_codex":[0.9925842,0.00016304,0.002384943,0.001422069,0.001748465,0.001697329],"domain_scores_gemma":[0.9934999,0.001889976,0.000932148,0.001810005,0.001194141,0.0006738132],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005858057,0.0002594229,0.0009889157,0.0003767233,0.0002384381,0.00004212482,0.004175997,0.0009501954,0.0005373559,0.1816707,0.143247,0.6669273],"study_design_scores_gemma":[0.0006782238,0.0009051459,0.0003112922,0.0008222525,0.0002120909,0.0003720804,0.0016247,0.7035728,0.00552177,0.0200474,0.2649595,0.0009727588],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02308889,0.002260099,0.9488914,0.01515468,0.0004096971,0.004367182,0.001720374,0.001474969,0.002632713],"genre_scores_gemma":[0.819076,0.0001867014,0.1717332,0.002982941,0.0005039081,0.003016636,0.0002201158,0.0001324625,0.00214802],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7959871,"threshold_uncertainty_score":0.9999522,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03255876441735268,"score_gpt":0.2927016363932953,"score_spread":0.2601428719759427,"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."}}