{"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/stochastic-answer-networks-for-machine","title":"Stochastic Answer Networks for Machine Reading Comprehension","arxiv_id":"1712.03556","date":"2017-12-10","proceeding":"ACL 2018 7","authors":["Xiaodong Liu","Yelong Shen","Kevin Duh","Jianfeng Gao"],"abstract":"We propose a simple yet robust stochastic answer network (SAN) that simulates\nmulti-step reasoning in machine reading comprehension. Compared to previous\nwork such as ReasoNet which used reinforcement learning to determine the number\nof steps, the unique feature is the use of a kind of stochastic prediction\ndropout on the answer module (final layer) of the neural network during the\ntraining. We show that this simple trick improves robustness and achieves\nresults competitive to the state-of-the-art on the Stanford Question Answering\nDataset (SQuAD), the Adversarial SQuAD, and the Microsoft MAchine Reading\nCOmprehension Dataset (MS MARCO).","url_abs":"http://arxiv.org/abs/1712.03556v2","url_pdf":"http://arxiv.org/pdf/1712.03556v2.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":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/Xaniar87/SAN_SQuAD2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/kevinduh/san_mrc","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/lduml/blog","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/om00839/machine-suneung","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/yongbowin/san_mrc_annotation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"stochastic-answer-networks-for-machine","repo_url":"https://github.com/MindCode-4/code-13/tree/main/stochastic-attention-head-removal-a-simple","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"machine-reading-comprehension","task_name":"Machine Reading Comprehension"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"reading-comprehension","task_name":"Reading Comprehension"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"SAN (ensemble model)","rank_in_archive_order":71,"of":213,"metrics":{"EM":"79.608","F1":"86.496"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"SAN (single model)","rank_in_archive_order":106,"of":213,"metrics":{"EM":"76.828","F1":"84.396"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11-dev","task":"Question Answering","dataset":"SQuAD1.1 dev","model":"SAN (single)","rank_in_archive_order":24,"of":55,"metrics":{"EM":"76.235","F1":"84.056"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad20","task":"Question Answering","dataset":"SQuAD2.0","model":"SAN (ensemble model)","rank_in_archive_order":249,"of":286,"metrics":{"EM":"71.316","F1":"73.704"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad20","task":"Question Answering","dataset":"SQuAD2.0","model":"SAN (single model)","rank_in_archive_order":257,"of":286,"metrics":{"EM":"68.653","F1":"71.439"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.03556","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}