{"id":"W3102559011","doi":"10.48550/arxiv.2011.06485","title":"Fairness and Robustness in Invariant Learning: A Case Study in Toxicity Classification","year":2020,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Robustness (evolution); Computer science; Machine learning; Artificial intelligence; Empirical risk minimization; Invariant (physics); Minification; Training set; Support vector machine; Generalization; Mathematics","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"],"consensus_categories":[],"category_scores_codex":[0.0006035332,0.000242755,0.000338257,0.0004397245,0.0001185729,0.000186452,0.000680235,0.0001900305,0.00000778229],"category_scores_gemma":[0.0001245665,0.0003014362,0.00004911648,0.001088705,0.00006069694,0.0005145029,0.0012671,0.001097485,0.000008657938],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002107432,"about_ca_system_score_gemma":0.0001773512,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0008776178,"about_ca_topic_score_gemma":0.001372308,"domain_scores_codex":[0.9975429,0.0006689123,0.0002774267,0.001163295,0.00009785104,0.0002496811],"domain_scores_gemma":[0.9989217,0.0001529125,0.0002310181,0.0004785744,0.00006779451,0.0001479804],"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":[0.00004310228,0.0003288591,0.1018149,0.00006563256,0.00002347784,0.01772323,0.006838325,0.8312566,0.00004756196,0.04028644,0.000004295555,0.001567514],"study_design_scores_gemma":[0.000989989,0.00007307712,0.0428775,0.00004674631,0.00001304011,0.0000587148,0.009614337,0.9446167,0.000004277832,0.001369257,0.00002761882,0.0003087891],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6567787,0.0000100743,0.3422453,0.0001590815,0.00008124978,0.0003420869,5.696654e-7,0.00008966608,0.0002932932],"genre_scores_gemma":[0.9989588,0.00002564017,0.000787523,0.00004556341,0.00001677647,0.000003434454,0.000004275866,0.00001257118,0.0001454458],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3421801,"threshold_uncertainty_score":0.9999438,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1633357046334301,"score_gpt":0.2266039780746313,"score_spread":0.06326827344120117,"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."}}