{"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/compare-compress-and-propagate-enhancing","title":"Compare, Compress and Propagate: Enhancing Neural Architectures with Alignment Factorization for Natural Language Inference","arxiv_id":"1801.00102","date":"2017-12-30","proceeding":"EMNLP 2018 10","authors":["Yi Tay","Luu Anh Tuan","Siu Cheung Hui"],"abstract":"This paper presents a new deep learning architecture for Natural Language\nInference (NLI). Firstly, we introduce a new architecture where alignment pairs\nare compared, compressed and then propagated to upper layers for enhanced\nrepresentation learning. Secondly, we adopt factorization layers for efficient\nand expressive compression of alignment vectors into scalar features, which are\nthen used to augment the base word representations. The design of our approach\nis aimed to be conceptually simple, compact and yet powerful. We conduct\nexperiments on three popular benchmarks, SNLI, MultiNLI and SciTail, achieving\ncompetitive performance on all. A lightweight parameterization of our model\nalso enjoys a $\\approx 3$ times reduction in parameter size compared to the\nexisting state-of-the-art models, e.g., ESIM and DIIN, while maintaining\ncompetitive performance. Additionally, visual analysis shows that our\npropagated features are highly interpretable.","url_abs":"http://arxiv.org/abs/1801.00102v2","url_pdf":"http://arxiv.org/pdf/1801.00102v2.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":[],"tasks":[{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"representation-learning","task_name":"Representation Learning"}],"methods":[{"method_slug":"esim","method_name":"ESIM"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"300D CAFE Ensemble","rank_in_archive_order":20,"of":98,"metrics":{"% Test Accuracy":"89.3","% Train Accuracy":"92.5","Parameters":"17.5m"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"300D CAFE","rank_in_archive_order":36,"of":98,"metrics":{"% Test Accuracy":"88.5","% Train Accuracy":"89.8","Parameters":"4.7m"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"300D CAFE (no cross-sentence attention)","rank_in_archive_order":65,"of":98,"metrics":{"% Test Accuracy":"85.9","% Train Accuracy":"87.3","Parameters":"3.7m"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-scitail","task":"Natural Language Inference","dataset":"SciTail","model":"CAFE","rank_in_archive_order":7,"of":13,"metrics":{"Accuracy":"83.3"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.00102","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}