{"id":"W6910401483","doi":"10.48448/ws9f-7304","title":"Debiasing should be Good and Bad: Measuring the Consistency of Debiasing Techniques in Language Models","year":2022,"lang":"en","type":"other","venue":"Open MIND","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Brock University","funders":"","keywords":"Debiasing; Consistency (knowledge bases); Interpretability; Generalizability theory; Protocol (science)","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":[0.07394627,0.0008004214,0.0006427406,0.003406664,0.001532384,0.005147922,0.002328287,0.003535671,0.001552273],"category_scores_gemma":[0.3873208,0.001024561,0.0008475992,0.002000588,0.004526706,0.009422011,0.006642381,0.003855719,0.0006212303],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002783521,"about_ca_system_score_gemma":0.002992402,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001993753,"about_ca_topic_score_gemma":0.001989323,"domain_scores_codex":[0.9081524,0.0436328,0.01208292,0.006789264,0.02740486,0.001937803],"domain_scores_gemma":[0.5046859,0.3058876,0.05272823,0.08976837,0.04408051,0.002849301],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.004496722,0.00301377,0.2963527,0.002281718,0.001115281,0.0008728966,0.04219545,0.04660973,0.08327011,0.08901656,0.009714822,0.4210601],"study_design_scores_gemma":[0.0007753284,0.007316142,0.1530004,0.001664458,0.0009345635,0.001560526,0.01800144,0.3414575,0.2777275,0.1561376,0.0403003,0.001124225],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6505526,0.0004757208,0.3304916,0.001948657,0.0001342506,0.002313735,0.001134409,0.003613464,0.009335648],"genre_scores_gemma":[0.8395394,0.0001307405,0.1555856,0.0004424413,0.00003396273,0.001412389,0.001002886,0.000606364,0.001246295],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.07394627,"threshold_uncertainty_score":0.3910699,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1140783712795055,"score_gpt":0.3390471434022081,"score_spread":0.2249687721227026,"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."}}