{"id":"W3174726724","doi":"10.18653/v1/2021.acl-long.163","title":"Optimizing Deeper Transformers on Small Datasets","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":57,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; McGill University; Canadian Institute for Advanced Research; University of Alberta; University of Waterloo","funders":"","keywords":"Computer science; Computational linguistics; Natural language processing; Transformer; Artificial intelligence; Library science; Cognitive science; Engineering; Psychology; Electrical engineering","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":[0.003729592,0.002238811,0.002996166,0.002090808,0.001348748,0.002988059,0.002964236,0.00158066,0.01422625],"category_scores_gemma":[0.01797511,0.001364457,0.002528004,0.003431226,0.001298704,0.01329503,0.003909285,0.003273311,0.003569372],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002296692,"about_ca_system_score_gemma":0.004132585,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0092453,"about_ca_topic_score_gemma":0.03209251,"domain_scores_codex":[0.9969435,0.0008505189,0.0002630802,0.0009592454,0.000517633,0.000465877],"domain_scores_gemma":[0.9910102,0.005139216,0.0002193006,0.002602858,0.000772402,0.0002560946],"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":[0.001775751,0.0008297619,0.008462082,0.0008589723,0.0004566411,0.000298609,0.0005178974,0.1552393,0.01134267,0.05194651,0.100802,0.6674699],"study_design_scores_gemma":[0.000309218,0.0001950223,0.0009575213,0.00006141548,0.0001699826,0.0001241203,0.0002897667,0.8614711,0.004820515,0.1253968,0.006181609,0.00002302628],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2543817,0.005256239,0.650514,0.003146841,0.0008041891,0.000723149,0.009363484,0.05892189,0.0168886],"genre_scores_gemma":[0.6183723,0.0008286446,0.344199,0.000834816,0.000287657,0.0003385564,0.020949,0.004556961,0.009633103],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01422625,"threshold_uncertainty_score":0.04759157,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03922932592945792,"score_gpt":0.2517262054989651,"score_spread":0.2124968795695072,"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."}}