{"id":"W4401042851","doi":"10.18653/v1/2024.findings-naacl.101","title":"TagDebias: Entity and Concept Tagging for Social Bias Mitigation in Pretrained Language Models","year":2024,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"Alliance de recherche numérique du Canada; Mitacs","keywords":"Computer science; Natural language processing; Artificial intelligence; Language model","routes":{"ca_aff":true,"ca_fund":true,"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":[0.00535435,0.002093633,0.001110892,0.001813398,0.001034499,0.001953264,0.002478997,0.002168545,0.003160498],"category_scores_gemma":[0.01398057,0.0006698259,0.001253614,0.001379012,0.0009156003,0.004340651,0.002884901,0.00391073,0.003688794],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009741551,"about_ca_system_score_gemma":0.001950409,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004081567,"about_ca_topic_score_gemma":0.01139274,"domain_scores_codex":[0.997552,0.001083161,0.0001400338,0.0006603625,0.0003558718,0.0002085304],"domain_scores_gemma":[0.9934506,0.003592572,0.0003348875,0.001625655,0.0007860643,0.0002101378],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00118446,0.0008438245,0.01511323,0.0006668804,0.0005997944,0.0005087623,0.0009363011,0.09524331,0.02556176,0.01307784,0.06196479,0.784299],"study_design_scores_gemma":[0.0001205299,0.0003029276,0.002366045,0.0001125591,0.0001260171,0.0003155758,0.0002717902,0.9356335,0.01728768,0.0267448,0.01662477,0.00009376217],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0880203,0.001995297,0.8833852,0.001276472,0.0009509845,0.0005257212,0.003251131,0.01656151,0.004033323],"genre_scores_gemma":[0.5490528,0.0008787638,0.4181583,0.002152191,0.0005580952,0.000968907,0.01535193,0.001550144,0.01132882],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.00535435,"threshold_uncertainty_score":0.0283168,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0276593721056513,"score_gpt":0.3086277423095566,"score_spread":0.2809683702039053,"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."}}