{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.005394307,0.0001476588,0.0003451964,0.00007537611,0.0004036052,0.00003345595,0.001624227,0.00007648215,4.11753e-7],"category_scores_gemma":[0.01372396,0.0001093241,0.0002679704,0.000492439,0.00007673676,0.00009501509,0.0005600764,0.0001808362,1.964249e-8],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002134226,"about_ca_system_score_gemma":0.0002106356,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001820424,"about_ca_topic_score_gemma":0.0000109515,"domain_scores_codex":[0.9977376,0.0002229808,0.0006522528,0.0002927287,0.0008750895,0.0002192884],"domain_scores_gemma":[0.992794,0.002299048,0.001655458,0.0002269359,0.002994149,0.0000304265],"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":[0.0003577547,0.0003859891,0.339095,0.001011758,0.0002224145,4.817702e-8,0.007939854,0.6044462,0.007893903,0.03764153,0.0002074001,0.0007982046],"study_design_scores_gemma":[0.0005996833,0.0002208614,0.1837891,0.0002806602,0.00009192355,0.000001217383,0.0003426751,0.7705804,0.01595123,0.02789908,0.00005176104,0.0001913972],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9236261,0.00005785748,0.07134483,0.0003178588,0.0009879509,0.002008164,0.0007883404,0.00008485106,0.0007840162],"genre_scores_gemma":[0.9693052,2.738809e-7,0.03041768,0.00005368645,0.0000792479,0.00005148637,0.00001178761,0.00001949102,0.00006113094],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1661342,"threshold_uncertainty_score":0.9945838,"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."}}