{"id":"W4400676406","doi":"10.1007/978-3-031-65112-0_6","title":"A Preliminary Study to Examining Per-class Performance Bias via Robustness Distributions","year":2024,"lang":"en","type":"book-chapter","venue":"Lecture notes in computer science","topic":"Imbalanced Data Classification Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Robustness (evolution); Computer science; Class (philosophy); Algorithm; Artificial intelligence","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","scholarly_communication","open_science"],"consensus_categories":[],"category_scores_codex":[0.001457216,0.0007099869,0.0006070081,0.001391387,0.000446769,0.001075219,0.005502158,0.0002455192,0.00001732983],"category_scores_gemma":[0.0001231564,0.0006624243,0.00009485523,0.001597696,0.0004696818,0.001058959,0.00391506,0.001168061,0.0002010749],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007356129,"about_ca_system_score_gemma":0.0005186526,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001476405,"about_ca_topic_score_gemma":0.00003132858,"domain_scores_codex":[0.9945155,0.00005572478,0.0007885526,0.002513608,0.001296028,0.0008305458],"domain_scores_gemma":[0.9957989,0.0004270233,0.0002686543,0.002859924,0.0003816694,0.0002637811],"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":[0.00002377071,0.000354668,0.0005689278,0.0001819456,0.000042302,0.0002421487,0.003892472,0.04930615,0.0003564246,0.01489337,0.000283533,0.9298543],"study_design_scores_gemma":[0.0002049921,0.001791898,0.002708073,0.0009178537,0.0000305211,0.0001792188,0.000002434625,0.9766061,0.002294192,0.0113656,0.002536096,0.001363016],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001931222,0.0001456676,0.9920763,0.0008382908,0.0016168,0.001268126,0.00003661567,0.0008848428,0.001202125],"genre_scores_gemma":[0.6940318,0.00001388614,0.3043465,0.0003807601,0.0002924801,0.000119851,0.00002921117,0.00005451426,0.0007309714],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9284913,"threshold_uncertainty_score":0.9999617,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04316998529932985,"score_gpt":0.2785133695285633,"score_spread":0.2353433842292335,"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."}}