{"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/sentence-encoders-on-stilts-supplementary","title":"Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks","arxiv_id":"1811.01088","date":"2018-11-02","proceeding":null,"authors":["Jason Phang","Thibault Févry","Samuel R. Bowman"],"abstract":"Pretraining sentence encoders with language modeling and related unsupervised\ntasks has recently been shown to be very effective for language understanding\ntasks. By supplementing language model-style pretraining with further training\non data-rich supervised tasks, such as natural language inference, we obtain\nadditional performance improvements on the GLUE benchmark. Applying\nsupplementary training on BERT (Devlin et al., 2018), we attain a GLUE score of\n81.8---the state of the art (as of 02/24/2019) and a 1.4 point improvement over\nBERT. We also observe reduced variance across random restarts in this setting.\nOur approach yields similar improvements when applied to ELMo (Peters et al.,\n2018a) and Radford et al. (2018)'s model. In addition, the benefits of\nsupplementary training are particularly pronounced in data-constrained regimes,\nas we show in experiments with artificially limited training data.","url_abs":"http://arxiv.org/abs/1811.01088v2","url_pdf":"http://arxiv.org/pdf/1811.01088v2.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":[{"paper_slug":"sentence-encoders-on-stilts-supplementary","repo_url":"https://github.com/zphang/bert_on_stilts","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[{"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":"bilstm","method_name":"BiLSTM"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"elmo","method_name":"ELMo"},{"method_slug":"lstm","method_name":"LSTM"},{"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":"residual-connection","method_name":"Residual Connection"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"},{"method_slug":"weight-decay","method_name":"Weight Decay"},{"method_slug":"wordpiece","method_name":"WordPiece"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1811.01088","atlas_url":"https://app.syntology.ai/?focus=1811.01088","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}