# A Light Recipe to Train Robust Vision Transformers

arXiv: https://arxiv.org/abs/2209.07399 · HTML: https://mlsanity.com/paper/2209.07399/

## Key facts

- Cited by 64 papers in the map (data through 14 Oct 2026).
- Builds on 60 of its 89 references that are in the map.
- First posted on arXiv 15 Sep 2022.

## Builds on (most cited first, 30 of 60)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Towards Deep Learning Models Resistant to Adversarial Attacks](https://mlsanity.com/paper/1706.06083/) (2017)
- [mixup: Beyond Empirical Risk Minimization](https://mlsanity.com/paper/1710.09412/) (2017)
- [Intriguing properties of neural networks](https://mlsanity.com/paper/1312.6199/) (2013)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Neural Machine Translation by Jointly Learning to Align and Translate](https://mlsanity.com/paper/1409.0473/) (2014)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [Gaussian Error Linear Units (GELUs)](https://mlsanity.com/paper/1606.08415/) (2016)
- [CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features](https://mlsanity.com/paper/1905.04899/) (2019)
- [BEiT: BERT Pre-Training of Image Transformers](https://mlsanity.com/paper/2106.08254/) (2021)
- [RandAugment: Practical automated data augmentation with a reduced search space](https://mlsanity.com/paper/1909.13719/) (2019)
- [Theoretically Principled Trade-off between Robustness and Accuracy](https://mlsanity.com/paper/1901.08573/) (2019)
- [Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks](https://mlsanity.com/paper/2003.01690/) (2020)
- [Random Erasing Data Augmentation](https://mlsanity.com/paper/1708.04896/) (2017)
- [Transformers in Vision: A Survey](https://mlsanity.com/paper/2101.01169/) (2021)
- [Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learning](https://mlsanity.com/paper/1702.03118/) (2017)
- [Revisiting Unreasonable Effectiveness of Data in Deep Learning Era](https://mlsanity.com/paper/1707.02968/) (2017)
- [Robustness May Be at Odds with Accuracy](https://mlsanity.com/paper/1805.12152/) (2018)
- [Going deeper with Image Transformers](https://mlsanity.com/paper/2103.17239/) (2021)
- [Fast is better than free: Revisiting adversarial training](https://mlsanity.com/paper/2001.03994/) (2020)
- [Evasion Attacks against Machine Learning at Test Time](https://mlsanity.com/paper/1708.06131/) (2017)
- [Square Attack: a query-efficient black-box adversarial attack via random search](https://mlsanity.com/paper/1912.00049/) (2019)
- [RobustBench: a standardized adversarial robustness benchmark](https://mlsanity.com/paper/2010.09670/) (2020)
- [Adversarial Training for Free!](https://mlsanity.com/paper/1904.12843/) (2019)
- [How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers](https://mlsanity.com/paper/2106.10270/) (2021)
- [MetaFormer Is Actually What You Need for Vision](https://mlsanity.com/paper/2111.11418/) (2021)

## Built on by (most cited first, 30 of 64)

- [Rethinking Model Ensemble in Transfer-based Adversarial Attacks](https://mlsanity.com/paper/2303.09105/) (2023)
- [A Comprehensive Study on Robustness of Image Classification Models: Benchmarking and Rethinking](https://mlsanity.com/paper/2302.14301/) (2023)
- [Revisiting Adversarial Training for ImageNet: Architectures, Training and Generalization across Threat Models](https://mlsanity.com/paper/2303.01870/) (2023)
- [Robust Classification via a Single Diffusion Model](https://mlsanity.com/paper/2305.15241/) (2023)
- [Robust Principles: Architectural Design Principles for Adversarially Robust CNNs](https://mlsanity.com/paper/2308.16258/) (2023)
- [Revisiting Adversarial Training at Scale](https://mlsanity.com/paper/2401.04727/) (2024)
- [ADBM: Adversarial diffusion bridge model for reliable adversarial purification](https://mlsanity.com/paper/2408.00315/) (2024)
- [Language-Driven Anchors for Zero-Shot Adversarial Robustness](https://mlsanity.com/paper/2301.13096/) (2023)
- [Revisiting adapters with adversarial training](https://mlsanity.com/paper/2210.04886/) (2022)
- [Improving the Accuracy-Robustness Trade-Off of Classifiers via Adaptive Smoothing](https://mlsanity.com/paper/2301.12554/) (2023)
- [MimicDiffusion: Purifying Adversarial Perturbation via Mimicking Clean Diffusion Model](https://mlsanity.com/paper/2312.04802/) (2023)
- [MixedNUTS: Training-Free Accuracy-Robustness Balance via Nonlinearly Mixed Classifiers](https://mlsanity.com/paper/2402.02263/) (2024)
- [AutoAdvExBench: Benchmarking autonomous exploitation of adversarial example defenses](https://mlsanity.com/paper/2503.01811/) (2025)
- [AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples](https://mlsanity.com/paper/2404.19460/) (2024)
- [On the Importance of Backbone to the Adversarial Robustness of Object Detectors](https://mlsanity.com/paper/2305.17438/) (2023)
- [Towards Reliable Evaluation and Fast Training of Robust Semantic Segmentation Models](https://mlsanity.com/paper/2306.12941/) (2023)
- [MIMIR: Masked Image Modeling for Mutual Information-based Adversarial Robustness](https://mlsanity.com/paper/2312.04960/) (2023)
- [Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers](https://mlsanity.com/paper/2406.18451/) (2024)
- [An Extended Study of Human-like Behavior under Adversarial Training](https://mlsanity.com/paper/2303.12669/) (2023)
- [FullLoRA: Efficiently Boosting the Robustness of Pretrained Vision Transformers](https://mlsanity.com/paper/2401.01752/) (2024)
- [On the Robustness of Dataset Inference](https://mlsanity.com/paper/2210.13631/) (2022)
- [Adversarial Guided Diffusion Models for Adversarial Purification](https://mlsanity.com/paper/2403.16067/) (2024)
- [σ-zero: Gradient-based Optimization of ℓ0-norm Adversarial Examples](https://mlsanity.com/paper/2402.01879/) (2024)
- [PartImageNet++ Dataset: Scaling up Part-based Models for Robust Recognition](https://mlsanity.com/paper/2407.10918/) (2024)
- [RobArch: Designing Robust Architectures against Adversarial Attacks](https://mlsanity.com/paper/2301.03110/) (2023)
- [Foundation Model-oriented Robustness: Robust Image Model Evaluation with Pretrained Models](https://mlsanity.com/paper/2308.10632/) (2023)
- [Adversarial Training: A Survey](https://mlsanity.com/paper/2410.15042/) (2024)
- [Class-Conditioned Transformation for Enhanced Robust Image Classification](https://mlsanity.com/paper/2303.15409/) (2023)
- [Segment (Almost) Nothing: Prompt-Agnostic Adversarial Attacks on Segmentation Models](https://mlsanity.com/paper/2311.14450/) (2023)
- [SpecFormer: Guarding Vision Transformer Robustness via Maximum Singular Value Penalization](https://mlsanity.com/paper/2402.03317/) (2024)

## Abstract (arXiv)

In this paper, we ask whether Vision Transformers (ViTs) can serve as an underlying architecture for improving the adversarial robustness of machine learning models against evasion attacks. While earlier works have focused on improving Convolutional Neural Networks, we show that also ViTs are highly suitable for adversarial training to achieve competitive performance. We achieve this objective using a custom adversarial training recipe, discovered using rigorous ablation studies on a subset of the ImageNet dataset. The canonical training recipe for ViTs recommends strong data augmentation, in part to compensate for the lack of vision inductive bias of attention modules, when compared to convolutions. We show that this recipe achieves suboptimal performance when used for adversarial training. In contrast, we find that omitting all heavy data augmentation, and adding some additional bag-of-tricks (varepsilon-warmup and larger weight decay), significantly boosts the performance of robust ViTs. We show that our recipe generalizes to different classes of ViT architectures and large-scale models on full ImageNet-1k. Additionally, investigating the reasons for the robustness of our models, we show that it is easier to generate strong attacks during training when using our recipe and that this leads to better robustness at test time. Finally, we further study one consequence of adversarial training by proposing a way to quantify the semantic nature of adversarial perturbations and highlight its correlation with the robustness of the model. Overall, we recommend that the community should avoid translating the canonical training recipes in ViTs to robust training and rethink common training choices in the context of adversarial training.