{"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/twin-networks-matching-the-future-for","title":"Twin Networks: Matching the Future for Sequence Generation","arxiv_id":"1708.06742","date":"2017-08-22","proceeding":"ICLR 2018 1","authors":["Dmitriy Serdyuk","Nan Rosemary Ke","Alessandro Sordoni","Adam Trischler","Chris Pal","Yoshua Bengio"],"abstract":"We propose a simple technique for encouraging generative RNNs to plan ahead.\nWe train a \"backward\" recurrent network to generate a given sequence in reverse\norder, and we encourage states of the forward model to predict cotemporal\nstates of the backward model. The backward network is used only during\ntraining, and plays no role during sampling or inference. We hypothesize that\nour approach eases modeling of long-term dependencies by implicitly forcing the\nforward states to hold information about the longer-term future (as contained\nin the backward states). We show empirically that our approach achieves 9%\nrelative improvement for a speech recognition task, and achieves significant\nimprovement on a COCO caption generation task.","url_abs":"http://arxiv.org/abs/1708.06742v3","url_pdf":"http://arxiv.org/pdf/1708.06742v3.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":"twin-networks-matching-the-future-for","repo_url":"https://github.com/dmitriy-serdyuk/twin-net","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"twin-networks-matching-the-future-for","repo_url":"https://github.com/mravanelli/pytorch-kaldi","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"caption-generation","task_name":"Caption Generation"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"},{"task_slug":"speech-recognition-1","task_name":"speech-recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1708.06742","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}