{"id":"W6929071316","doi":"10.48448/a98m-1a69","title":"End-to-End Self-Debiasing Framework for Robust NLU Training","year":2021,"lang":"en","type":"other","venue":"Open MIND","topic":"Chemical and Physical Studies","field":"Biochemistry, Genetics and Molecular Biology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; Université de Montréal","funders":"","keywords":"Debiasing; Natural language understanding; Training (meteorology); Simple (philosophy); Training set; Task (project management)","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":[0.001722067,0.001615974,0.001250732,0.0009044432,0.0008374733,0.001374506,0.003158232,0.002622778,0.01288681],"category_scores_gemma":[0.00555874,0.0007229525,0.0009226042,0.0006917564,0.0009868658,0.002604155,0.003782441,0.003229323,0.00798092],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008858391,"about_ca_system_score_gemma":0.001495665,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002955529,"about_ca_topic_score_gemma":0.006137984,"domain_scores_codex":[0.9992363,0.0001514897,0.00004872062,0.000315131,0.000161677,0.00008667004],"domain_scores_gemma":[0.9983751,0.0005850347,0.0001055733,0.0005045561,0.0003466881,0.00008305552],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003672804,0.0002408627,0.001217275,0.0002769309,0.00008878906,0.0003015998,0.0003107439,0.09387717,0.02917367,0.01012584,0.01942095,0.8445989],"study_design_scores_gemma":[0.00002882434,0.00008010266,0.000246973,0.00002378213,0.00001634843,0.0001266623,0.00006862399,0.951614,0.02629927,0.01375557,0.007714687,0.00002514873],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01145046,0.0003407745,0.9499207,0.0001896138,0.00009131847,0.0001439258,0.0004166942,0.03480552,0.002640968],"genre_scores_gemma":[0.2005211,0.0002253873,0.7795147,0.0006522337,0.00008478142,0.0004875913,0.003117207,0.002925423,0.01247157],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01288681,"threshold_uncertainty_score":0.04311073,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0669392359194059,"score_gpt":0.3259969365856715,"score_spread":0.2590577006662657,"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."}}