mlsanityData to 7 Oct 2026

Quadratic Upper Bound for Boosting Robustness

arXiv 2601.13645 ↗First posted 20 Jan 2026cs.LG cs.AI cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Fast adversarial training (FAT) aims to enhance the robustness of models against adversarial attacks with reduced training time, however, FAT often suffers from compromised robustness due to insufficient exploration of adversarial space. In this paper, we develop a loss function to mitigate the problem of degraded robustness under FAT. Specifically, we derive a quadratic upper bound (QUB) on the adversarial training (AT) loss function and propose to utilize the bound with existing FAT methods. Our experimental results show that applying QUB loss to the existing methods yields significant improvement of robustness. Furthermore, using various metrics, we demonstrate that this improvement is likely to result from the smoothened loss landscape of the resulting model.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 38 of its 45 references that are in the map. First posted on arXiv 20 Jan 2026.

Builds on

38 of 45 references.

References · 38

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  4. 2016Wide Residual Networkscited by 3,057
  5. 2016Identity Mappings in Deep Residual Networkscited by 1,960
  6. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
  7. 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
  8. 2016An overview of gradient descent optimization algorithmscited by 1,021
  9. 2016Adversarial Machine Learning at Scalecited by 934
  10. 2017Ensemble Adversarial Training: Attacks and Defensescited by 833
  11. 2020Fast is better than free: Revisiting adversarial trainingcited by 723
  12. 2019Square Attack: a query-efficient black-box adversarial attack via random searchcited by 682
26 more
  1. 2019Adversarial Training for Free!cited by 638
  2. 2020Overfitting in adversarially robust deep learningcited by 504
  3. 2019Minimally distorted Adversarial Examples with a Fast Adaptive Boundary Attackcited by 333
  4. 2020Understanding and Improving Fast Adversarial Trainingcited by 213
  5. 2020Adversarial Distributional Training for Robust Deep Learningcited by 79
  6. 2020Understanding Catastrophic Overfitting in Single-step Adversarial Trainingcited by 75
  7. 2020Guided Adversarial Attack for Evaluating and Enhancing Adversarial Defensescited by 61
  8. 2022Make Some Noise: Reliable and Efficient Single-Step Adversarial Trainingcited by 42
  9. 2021Subspace Adversarial Trainingcited by 41
  10. 2019Jacobian Adversarially Regularized Networks for Robustnesscited by 40
  11. 2020Single-step Adversarial training with Dropout Schedulingcited by 40
  12. 2021Boosting Fast Adversarial Training with Learnable Adversarial Initializationcited by 40
  13. 2022Prior-Guided Adversarial Initialization for Fast Adversarial Trainingcited by 34
  14. 2023Improving Fast Adversarial Training with Prior-Guided Knowledgecited by 34
  15. 2021Reliably fast adversarial training via latent adversarial perturbationcited by 29
  16. 2022Fast Adversarial Training with Adaptive Step Sizecited by 24
  17. 2022Understanding and Combating Robust Overfitting via Input Loss Landscape Analysis and Regularizationcited by 22
  18. 2023On the Over-Memorization During Natural, Robust and Catastrophic Overfittingcited by 19
  19. 2023Revisiting and Exploring Efficient Fast Adversarial Training via LAW: Lipschitz Regularization and Auto Weight Averagingcited by 14
  20. 2023Fast Adversarial Training with Smooth Convergencecited by 14
  21. 2021Understanding Catastrophic Overfitting in Adversarial Trainingcited by 11
  22. 2018Interpreting Adversarial Robustness: A View from Decision Surface in Input Spacecited by 10
  23. 2022How many perturbations break this model? Evaluating robustness beyond adversarial accuracycited by 9
  24. 2024Layer-Aware Analysis of Catastrophic Overfitting: Revealing the Pseudo-Robust Shortcut Dependencycited by 9
  25. 2019Lecture Notes: Optimization for Machine Learningcited by 8
  26. 2024Efficient local linearity regularization to overcome catastrophic overfittingcited by 7

Built on by

Not cited yet.