{"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/dcn-mixed-objective-and-deep-residual","title":"DCN+: Mixed Objective and Deep Residual Coattention for Question Answering","arxiv_id":"1711.00106","date":"2017-10-31","proceeding":"ICLR 2018 1","authors":["Caiming Xiong","Victor Zhong","Richard Socher"],"abstract":"Traditional models for question answering optimize using cross entropy loss,\nwhich encourages exact answers at the cost of penalizing nearby or overlapping\nanswers that are sometimes equally accurate. We propose a mixed objective that\ncombines cross entropy loss with self-critical policy learning. The objective\nuses rewards derived from word overlap to solve the misalignment between\nevaluation metric and optimization objective. In addition to the mixed\nobjective, we improve dynamic coattention networks (DCN) with a deep residual\ncoattention encoder that is inspired by recent work in deep self-attention and\nresidual networks. Our proposals improve model performance across question\ntypes and input lengths, especially for long questions that requires the\nability to capture long-term dependencies. On the Stanford Question Answering\nDataset, our model achieves state-of-the-art results with 75.1% exact match\naccuracy and 83.1% F1, while the ensemble obtains 78.9% exact match accuracy\nand 86.0% F1.","url_abs":"http://arxiv.org/abs/1711.00106v2","url_pdf":"http://arxiv.org/pdf/1711.00106v2.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":"dcn-mixed-objective-and-deep-residual","repo_url":"https://github.com/lmn-extracts/dcn_plus","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"DCN+ (ensemble)","rank_in_archive_order":81,"of":213,"metrics":{"EM":"78.852","F1":"85.996"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"DCN+ (single model)","rank_in_archive_order":125,"of":213,"metrics":{"EM":"74.866","F1":"82.806"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11-dev","task":"Question Answering","dataset":"SQuAD1.1 dev","model":"DCN+ (single)","rank_in_archive_order":29,"of":55,"metrics":{"EM":"74.5","F1":"83.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.00106","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}