{"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/reinforced-mnemonic-reader-for-machine","title":"Reinforced Mnemonic Reader for Machine Reading Comprehension","arxiv_id":"1705.02798","date":"2017-05-08","proceeding":null,"authors":["Minghao Hu","Yuxing Peng","Zhen Huang","Xipeng Qiu","Furu Wei","Ming Zhou"],"abstract":"In this paper, we introduce the Reinforced Mnemonic Reader for machine\nreading comprehension tasks, which enhances previous attentive readers in two\naspects. First, a reattention mechanism is proposed to refine current\nattentions by directly accessing to past attentions that are temporally\nmemorized in a multi-round alignment architecture, so as to avoid the problems\nof attention redundancy and attention deficiency. Second, a new optimization\napproach, called dynamic-critical reinforcement learning, is introduced to\nextend the standard supervised method. It always encourages to predict a more\nacceptable answer so as to address the convergence suppression problem occurred\nin traditional reinforcement learning algorithms. Extensive experiments on the\nStanford Question Answering Dataset (SQuAD) show that our model achieves\nstate-of-the-art results. Meanwhile, our model outperforms previous systems by\nover 6% in terms of both Exact Match and F1 metrics on two adversarial SQuAD\ndatasets.","url_abs":"http://arxiv.org/abs/1705.02798v6","url_pdf":"http://arxiv.org/pdf/1705.02798v6.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":"reinforced-mnemonic-reader-for-machine","repo_url":"https://github.com/HKUST-KnowComp/MnemonicReader","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"reinforced-mnemonic-reader-for-machine","repo_url":"https://github.com/ewrfcas/Reinforced-Mnemonic-Reader","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"reinforced-mnemonic-reader-for-machine","repo_url":"https://github.com/yly-revive/chainer-mreader","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","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":"Reinforced Mnemonic Reader (ensemble model)","rank_in_archive_order":45,"of":213,"metrics":{"EM":"82.283","F1":"88.533"},"uses_additional_data":true},{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"Reinforced Mnemonic Reader (single model)","rank_in_archive_order":73,"of":213,"metrics":{"EM":"79.545","F1":"86.654"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"Mnemonic Reader (ensemble)","rank_in_archive_order":129,"of":213,"metrics":{"EM":"74.268","F1":"82.371"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11","task":"Question Answering","dataset":"SQuAD1.1","model":"Mnemonic Reader (single model)","rank_in_archive_order":154,"of":213,"metrics":{"EM":"70.995","F1":"80.146"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11-dev","task":"Question Answering","dataset":"SQuAD1.1 dev","model":"R.M-Reader (single)","rank_in_archive_order":17,"of":55,"metrics":{"EM":"78.9","F1":" 86.3"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-triviaqa","task":"Question Answering","dataset":"TriviaQA","model":"Mnemonic Reader","rank_in_archive_order":45,"of":56,"metrics":{"EM":"46.94","F1":"52.85"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.02798","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}