Learning to Ignore Adversarial Attacks
Not on the map. Why?
Abstract (arXiv)
Despite the strong performance of current NLP models, they can be brittle against adversarial attacks. To enable effective learning against adversarial inputs, we introduce the use of rationale models that can explicitly learn to ignore attack tokens. We find that the rationale models can successfully ignore over 90% of attack tokens. This approach leads to consistent sizable improvements (~10%) over baseline models in robustness on three datasets for both BERT and RoBERTa, and also reliably outperforms data augmentation with adversarial examples alone. In many cases, we find that our method is able to close the gap between model performance on a clean test set and an attacked test set and hence reduce the effect of adversarial attacks.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 33 of its 50 references that are in the map. First posted on arXiv 23 May 2022.
Builds on
33 of 50 references.
References · 33
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
- 2013Intriguing properties of neural networkscited by 5,155
- 2016SQuAD: 100,000+ Questions for Machine Comprehension of Textcited by 4,148
- 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
- 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
- 2018FEVER: a large-scale dataset for Fact Extraction and VERificationcited by 1,375
- 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,340
- 2019Universal Adversarial Triggers for Attacking and Analyzing NLPcited by 722
- 2015Distillation as a Defense to Adversarial Perturbations against Deep Neural Networkscited by 678
- 2019Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailmentcited by 671
- 2017Adversarial Examples for Evaluating Reading Comprehension Systemscited by 571
21 more
- 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 530
- 2019Using Self-Supervised Learning Can Improve Model Robustness and Uncertaintycited by 514
- 2020BERT-ATTACK: Adversarial Attack Against BERT Using BERTcited by 491
- 2019On Evaluating Adversarial Robustnesscited by 447
- 2018Generating Natural Language Adversarial Examplescited by 424
- 2018TextBugger: Generating Adversarial Text Against Real-world Applicationscited by 381
- 2018Black-box Generation of Adversarial Text Sequences to Evade Deep Learning Classifierscited by 345
- 2020BAE: BERT-based Adversarial Examples for Text Classificationcited by 316
- 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 294
- 2016Rationalizing Neural Predictionscited by 281
- 2016Understanding Neural Networks through Representation Erasurecited by 224
- 2019Certified Robustness to Adversarial Word Substitutionscited by 168
- 2018Pathologies of Neural Models Make Interpretations Difficultcited by 140
- 2020Contextualized Perturbation for Textual Adversarial Attackcited by 136
- 2021Towards Improving Adversarial Training of NLP Modelscited by 79
- 2021When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Datacited by 75
- 2022Can Rationalization Improve Robustness?cited by 36
- 2020Defense against Adversarial Attacks in NLP via Dirichlet Neighborhood Ensemblecited by 28
- 2021Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updatescited by 27
- 2021Explainable Machine Learning with Prior Knowledge: An Overviewcited by 15
- 2021Practical Benefits of Feature Feedback Under Distribution Shiftcited by 5
Built on by
Citing papers per year: 2023: 1; 2024: 1.
2 papers.