Proceedings

Generalization Analysis for Adversarial Vision Transformer

Ziwen Jiang, Chang Cao, Han Li, Hong Chen, Rushi Lan

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Main Track. Pages 1279-1287.

https://doi.org/10.24963/ijcai.2026/143

Keywords

  • Machine Learning: Robustness
  • Computer Vision: Machine learning for vision
  • Machine Learning: Adversarial machine learning

Abstract

Vision Transformers (ViTs) exhibit notable susceptibility to adversarial attacks, presenting a significant challenge for their deployment in security-sensitive applications. Despite their considerable empirical successes, a rigorous theoretical foundation for ViT's adversarial generalization behavior has not been adequately established. To address this limitation, we leverage empirical Rademacher complexity to analyze the mechanism of perturbation accumulation through deep ViTs layers. We establish a high-probability generalization bound for ViTs in classification tasks under adversarial settings. Our theoretical framework elucidates the roles of several factors in mitigating perturbation effects, norm regularization of weight matrices (in both MLP and attention modules) and depth-wise propagation constraints on layer-wise norms. Extensive experiments on benchmark datasets corroborate our theoretical insights, bridging the gap between ViTs architecture design and adversarial robustness.

Ziwen Jiang, Chang Cao, Han Li, Hong Chen, and Rushi Lan. 2026. Generalization Analysis for Adversarial Vision Transformer. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 1279–1287. https://doi.org/10.24963/ijcai.2026/143
Welcome Future Conferences AI Journal Division IJCAI Proceedings About IJCAI IJCAI Awards