{"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/adversarially-regularising-neural-nli-models","title":"Adversarially Regularising Neural NLI Models to Integrate Logical Background Knowledge","arxiv_id":"1808.08609","date":"2018-08-26","proceeding":"CONLL 2018 10","authors":["Pasquale Minervini","Sebastian Riedel"],"abstract":"Adversarial examples are inputs to machine learning models designed to cause\nthe model to make a mistake. They are useful for understanding the shortcomings\nof machine learning models, interpreting their results, and for regularisation.\nIn NLP, however, most example generation strategies produce input text by using\nknown, pre-specified semantic transformations, requiring significant manual\neffort and in-depth understanding of the problem and domain. In this paper, we\ninvestigate the problem of automatically generating adversarial examples that\nviolate a set of given First-Order Logic constraints in Natural Language\nInference (NLI). We reduce the problem of identifying such adversarial examples\nto a combinatorial optimisation problem, by maximising a quantity measuring the\ndegree of violation of such constraints and by using a language model for\ngenerating linguistically-plausible examples. Furthermore, we propose a method\nfor adversarially regularising neural NLI models for incorporating background\nknowledge. Our results show that, while the proposed method does not always\nimprove results on the SNLI and MultiNLI datasets, it significantly and\nconsistently increases the predictive accuracy on adversarially-crafted\ndatasets -- up to a 79.6% relative improvement -- while drastically reducing\nthe number of background knowledge violations. Furthermore, we show that\nadversarial examples transfer among model architectures, and that the proposed\nadversarial training procedure improves the robustness of NLI models to\nadversarial examples.","url_abs":"http://arxiv.org/abs/1808.08609v1","url_pdf":"http://arxiv.org/pdf/1808.08609v1.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":"adversarially-regularising-neural-nli-models","repo_url":"https://github.com/uclmr/adversarial-nli","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"adversarially-regularising-neural-nli-models","repo_url":"https://github.com/makcedward/nlpaug","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1808.08609","atlas_url":"https://app.syntology.ai/?focus=1808.08609","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}