{"id":"W4385574336","doi":"10.18653/v1/2022.findings-emnlp.363","title":"Improving Generalization of Pre-trained Language Models via Stochastic Weight Averaging","year":2022,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; Université de Montréal","funders":"","keywords":"Generalization; Computer science; Language model; Artificial intelligence; Flatness (cosmology); Computation; Distillation; Machine learning; Convergence (economics); Algorithm; Mathematics","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":[0.002236412,0.002101206,0.00161387,0.0009844013,0.0005574418,0.001140139,0.001867248,0.001462235,0.001874428],"category_scores_gemma":[0.009707629,0.0009318556,0.001473566,0.0009576058,0.0008859361,0.003538914,0.002264009,0.00318115,0.001203348],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009224016,"about_ca_system_score_gemma":0.001339038,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007859498,"about_ca_topic_score_gemma":0.01262472,"domain_scores_codex":[0.9991043,0.0002950825,0.00007342481,0.0002807334,0.0001640966,0.00008238087],"domain_scores_gemma":[0.9968423,0.00195305,0.0001726391,0.000619868,0.0003263573,0.00008580295],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00008597302,0.0001111581,0.0008141215,0.0001194329,0.0001584837,0.0001020036,0.000170352,0.7970065,0.00938088,0.007282043,0.002729686,0.1820394],"study_design_scores_gemma":[0.000005445251,0.00002087034,0.00008632335,0.00000502178,0.00001007154,0.00001213494,0.000007535265,0.9948226,0.0009268888,0.003837839,0.0002592177,0.000006026757],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.04087031,0.0005892828,0.9540103,0.0002823457,0.00006333121,0.00004962762,0.0001382066,0.002834052,0.00116257],"genre_scores_gemma":[0.6984598,0.0008169333,0.2922725,0.0006441017,0.0001868419,0.0003306662,0.001631746,0.0009233901,0.004733938],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.007859498,"threshold_uncertainty_score":0.0156275,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0105345872943328,"score_gpt":0.2211012462258387,"score_spread":0.2105666589315059,"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."}}