# Robust Machine Comprehension Models via Adversarial Training

arXiv: https://arxiv.org/abs/1804.06473 · HTML: https://mlsanity.com/paper/1804.06473/

## Key facts

- Cited by 44 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Apr 2018.

## Built on by (most cited first, 30 of 44)

- [Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks](https://mlsanity.com/paper/2310.10844/) (2023)
- [An Empirical Survey of Data Augmentation for Limited Data Learning in NLP](https://mlsanity.com/paper/2106.07499/) (2021)
- [A Survey of Safety and Trustworthiness of Large Language Models through the Lens of Verification and Validation](https://mlsanity.com/paper/2305.11391/) (2023)
- [CLINE: Contrastive Learning with Semantic Negative Examples for Natural Language Understanding](https://mlsanity.com/paper/2107.00440/) (2021)
- [Can NLI Models Verify QA Systems' Predictions?](https://mlsanity.com/paper/2104.08731/) (2021)
- [HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalization](https://mlsanity.com/paper/2106.00149/) (2021)
- [A Survey of Adversarial Defences and Robustness in NLP](https://mlsanity.com/paper/2203.06414/) (2022)
- [Certified Robustness to Programmable Transformations in LSTMs](https://mlsanity.com/paper/2102.07818/) (2021)
- [Debiasing Masks: A New Framework for Shortcut Mitigation in NLU](https://mlsanity.com/paper/2210.16079/) (2022)
- [Recent Advances in Neural Text Generation: A Task-Agnostic Survey](https://mlsanity.com/paper/2203.03047/) (2022)
- [Contrastive Fine-tuning Improves Robustness for Neural Rankers](https://mlsanity.com/paper/2105.12932/) (2021)
- [Unifying Model Explainability and Robustness for Joint Text Classification and Rationale Extraction](https://mlsanity.com/paper/2112.10424/) (2021)
- [PIDformer: Transformer Meets Control Theory](https://mlsanity.com/paper/2402.15989/) (2024)
- [Improving Conversational Recommendation Systems via Counterfactual Data Simulation](https://mlsanity.com/paper/2306.02842/) (2023)
- [Model-tuning Via Prompts Makes NLP Models Adversarially Robust](https://mlsanity.com/paper/2303.07320/) (2023)
- [Residue-Based Natural Language Adversarial Attack Detection](https://mlsanity.com/paper/2204.10192/) (2022)
- [Improving the robustness and accuracy of biomedical language models through adversarial training](https://mlsanity.com/paper/2111.08529/) (2021)
- [Desiderata for the Context Use of Question Answering Systems](https://mlsanity.com/paper/2401.18001/) (2024)
- [Are Multilingual BERT models robust? A Case Study on Adversarial Attacks for Multilingual Question Answering](https://mlsanity.com/paper/2104.07646/) (2021)
- [LEAP: Efficient and Automated Test Method for NLP Software](https://mlsanity.com/paper/2308.11284/) (2023)
- [Virtual Data Augmentation: A Robust and General Framework for Fine-tuning Pre-trained Models](https://mlsanity.com/paper/2109.05793/) (2021)
- [Enhancing Model Robustness By Incorporating Adversarial Knowledge Into Semantic Representation](https://mlsanity.com/paper/2102.11584/) (2021)
- [Reasoning Chain Based Adversarial Attack for Multi-hop Question Answering](https://mlsanity.com/paper/2112.09658/) (2021)
- [A Survey on Measuring and Mitigating Reasoning Shortcuts in Machine Reading Comprehension](https://mlsanity.com/paper/2209.01824/) (2022)
- [The Impacts of Unanswerable Questions on the Robustness of Machine Reading Comprehension Models](https://mlsanity.com/paper/2302.00094/) (2023)
- [Towards Confident Machine Reading Comprehension](https://mlsanity.com/paper/2101.07942/) (2021)
- [Adversary for Social Good: Leveraging Adversarial Attacks to Protect Personal Attribute Privacy](https://mlsanity.com/paper/2306.02488/) (2023)
- [Model Agnostic Answer Reranking System for Adversarial Question Answering](https://mlsanity.com/paper/2102.03016/) (2021)
- [Improving Robustness of Malware Classifiers using Adversarial Strings Generated from Perturbed Latent Representations](https://mlsanity.com/paper/2110.11987/) (2021)
- [Data Augmentation for Neural NLP](https://mlsanity.com/paper/2302.11412/) (2023)

## Abstract (arXiv)

It is shown that many published models for the Stanford Question Answering Dataset (Rajpurkar et al., 2016) lack robustness, suffering an over 50% decrease in F1 score during adversarial evaluation based on the AddSent (Jia and Liang, 2017) algorithm. It has also been shown that retraining models on data generated by AddSent has limited effect on their robustness. We propose a novel alternative adversary-generation algorithm, AddSentDiverse, that significantly increases the variance within the adversarial training data by providing effective examples that punish the model for making certain superficial assumptions. Further, in order to improve robustness to AddSent's semantic perturbations (e.g., antonyms), we jointly improve the model's semantic-relationship learning capabilities in addition to our AddSentDiverse-based adversarial training data augmentation. With these additions, we show that we can make a state-of-the-art model significantly more robust, achieving a 36.5% increase in F1 score under many different types of adversarial evaluation while maintaining performance on the regular SQuAD task.