{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/cloze-driven-pretraining-of-self-attention","title":"Cloze-driven Pretraining of Self-attention Networks","arxiv_id":"1903.07785","date":"2019-03-19","proceeding":"IJCNLP 2019 11","authors":["Alexei Baevski","Sergey Edunov","Yinhan Liu","Luke Zettlemoyer","Michael Auli"],"abstract":"We present a new approach for pretraining a bi-directional transformer model\nthat provides significant performance gains across a variety of language\nunderstanding problems. Our model solves a cloze-style word reconstruction\ntask, where each word is ablated and must be predicted given the rest of the\ntext. Experiments demonstrate large performance gains on GLUE and new state of\nthe art results on NER as well as constituency parsing benchmarks, consistent\nwith the concurrently introduced BERT model. We also present a detailed\nanalysis of a number of factors that contribute to effective pretraining,\nincluding data domain and size, model capacity, and variations on the cloze\nobjective.","url_abs":"http://arxiv.org/abs/1903.07785v1","url_pdf":"http://arxiv.org/pdf/1903.07785v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"constituency-parsing","task_name":"Constituency Parsing"},{"task_slug":"cg","task_name":"NER"},{"task_slug":"named-entity-recognition-ner","task_name":"Named Entity Recognition (NER)"},{"task_slug":"sentiment-analysis","task_name":"Sentiment Analysis"},{"task_slug":"text-classification","task_name":"Text Classification"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"attention-dropout","method_name":"Attention Dropout"},{"method_slug":"bert","method_name":"BERT"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"linear-warmup-with-linear-decay","method_name":"Linear Warmup With Linear Decay"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"},{"method_slug":"weight-decay","method_name":"Weight Decay"},{"method_slug":"wordpiece","method_name":"WordPiece"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/constituency-parsing-on-penn-treebank","task":"Constituency Parsing","dataset":"Penn Treebank","model":"CNN Large + fine-tune","rank_in_archive_order":12,"of":27,"metrics":{"F1 score":"95.6"},"uses_additional_data":false},{"leaderboard":"/sota/named-entity-recognition-ner-on-conll-2003","task":"Named Entity Recognition (NER)","dataset":"CoNLL 2003 (English)","model":"CNN Large + fine-tune","rank_in_archive_order":18,"of":73,"metrics":{"F1":"93.5"},"uses_additional_data":true},{"leaderboard":"/sota/sentiment-analysis-on-sst-2-binary","task":"Sentiment Analysis","dataset":"SST-2 Binary classification","model":"CNN Large","rank_in_archive_order":33,"of":87,"metrics":{"Accuracy":"94.6"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1903.07785","atlas_url":"https://app.syntology.ai/?focus=1903.07785","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}