{"id":"W3206487987","doi":"10.18653/v1/2022.acl-long.132","title":"An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre-trained Language Models","year":2022,"lang":"en","type":"article","venue":"Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","topic":"Topic Modeling","field":"Computer Science","cited_by":109,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Institut de Valorisation des Données; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Debiasing; Counterfactual thinking; Computer science; Dropout (neural networks); Machine learning; Artificial intelligence; Psychology; Social psychology","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":[0.01878466,0.00238255,0.001267415,0.002218169,0.0009484035,0.001672335,0.002347504,0.001759715,0.00151298],"category_scores_gemma":[0.07769394,0.0008173544,0.001408673,0.001757824,0.001408585,0.004289914,0.002336908,0.003413656,0.00142536],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00132604,"about_ca_system_score_gemma":0.00161376,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004811323,"about_ca_topic_score_gemma":0.007090756,"domain_scores_codex":[0.9876691,0.00691467,0.001131275,0.001946485,0.001999673,0.000338836],"domain_scores_gemma":[0.9300296,0.05421005,0.002308039,0.008962282,0.003898314,0.0005916079],"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":[0.001062337,0.0004360542,0.03509329,0.002934292,0.001236878,0.0001364379,0.0009953333,0.08045357,0.01176903,0.003121842,0.01276632,0.8499946],"study_design_scores_gemma":[0.0002920426,0.002867921,0.03830665,0.002114162,0.001120418,0.001196973,0.001544815,0.8164682,0.08681882,0.01159484,0.03738175,0.0002935065],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.544589,0.07442861,0.3415494,0.004115991,0.0006947732,0.0006975526,0.002820278,0.01953557,0.01156888],"genre_scores_gemma":[0.7411374,0.01122036,0.2352809,0.0009108141,0.0002645931,0.0003962124,0.006678056,0.001481732,0.002629918],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01878466,"threshold_uncertainty_score":0.09934396,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01697158897121612,"score_gpt":0.2959718698235908,"score_spread":0.2790002808523747,"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."}}