{"id":"W3135261467","doi":"10.48550/arxiv.2103.03098","title":"Accounting for Variance in Machine Learning Benchmarks","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":40,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal; Canadian Institute for Advanced Research; McGill University; École de Technologie Supérieure; Université de Montréal","funders":"","keywords":"Variance (accounting); Computer science; Artificial intelligence; Accounting; Machine learning; Economics","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":[0.02161669,0.001593788,0.001587005,0.001664932,0.0009586849,0.002707478,0.002574245,0.00192272,0.00172252],"category_scores_gemma":[0.1309211,0.0005597568,0.001178096,0.002013631,0.002452894,0.003576319,0.002991475,0.003306382,0.0006225149],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001476323,"about_ca_system_score_gemma":0.001594823,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001738336,"about_ca_topic_score_gemma":0.00217164,"domain_scores_codex":[0.9766148,0.01305066,0.001279705,0.003004998,0.004960595,0.001089228],"domain_scores_gemma":[0.9178931,0.05382477,0.003743696,0.01998066,0.003788566,0.000769164],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001073907,0.0005093054,0.03319813,0.0007959215,0.0008632099,0.0002935434,0.0002779115,0.7854847,0.007548311,0.04325171,0.01535726,0.1113461],"study_design_scores_gemma":[0.00009325987,0.0005331562,0.007673598,0.0001525687,0.0001203399,0.0002287295,0.0001089855,0.9012634,0.01404359,0.07158127,0.004133468,0.0000676754],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.4215859,0.007272026,0.5444962,0.003838923,0.001180364,0.0004356739,0.002237712,0.006287106,0.01266608],"genre_scores_gemma":[0.9339749,0.0003655686,0.06082234,0.0004575944,0.0001746228,0.0002789078,0.001852126,0.0008410539,0.001233073],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02161669,"threshold_uncertainty_score":0.1143213,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04317978687297177,"score_gpt":0.2012533850704508,"score_spread":0.1580735981974791,"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."}}