{"id":"W4399156410","doi":"10.1145/3654934","title":"Data Acquisition for Improving Model Confidence","year":2024,"lang":"en","type":"article","venue":"Proceedings of the ACM on Management of Data","topic":"Machine Learning and Data Classification","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; York University","funders":"","keywords":"Computer science; Data acquisition; Machine learning; Context (archaeology); Knowledge acquisition; Range (aeronautics); Process (computing); Artificial intelligence; Data mining; Data quality; Data science; Engineering","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":[0.01375825,0.002195951,0.002893644,0.003292915,0.001276069,0.003068318,0.004153091,0.002334895,0.003508053],"category_scores_gemma":[0.09475756,0.001248984,0.00176109,0.002821975,0.001729842,0.008389064,0.006667011,0.005406511,0.001921993],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0011821,"about_ca_system_score_gemma":0.00405518,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003463033,"about_ca_topic_score_gemma":0.005274174,"domain_scores_codex":[0.9912096,0.002772268,0.0009594766,0.001963483,0.002614927,0.0004802168],"domain_scores_gemma":[0.9360967,0.03444773,0.004057187,0.01517976,0.009325502,0.0008930379],"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":[0.00161165,0.001265085,0.02763243,0.0008276705,0.0004181537,0.0003846941,0.001237165,0.1730256,0.0189517,0.02035347,0.01468798,0.7396045],"study_design_scores_gemma":[0.0001074617,0.0003580607,0.002721831,0.00009817799,0.00007984424,0.0002477498,0.0002827177,0.9482003,0.01522121,0.0268307,0.005794269,0.00005767327],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04688839,0.0008294734,0.9459043,0.0007758488,0.00007480355,0.0002772324,0.0006579277,0.003196,0.001396068],"genre_scores_gemma":[0.4180884,0.0004560025,0.5730137,0.0008035888,0.0001673211,0.0006793083,0.004735751,0.0007837301,0.001272218],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01375825,"threshold_uncertainty_score":0.07276142,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1126036958320424,"score_gpt":0.3468254042837654,"score_spread":0.234221708451723,"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."}}