Research map / Adversarial robustness and security / Certified and empirical robustness
Regional Adversarial Training for Better Robust Generalization
Chuanbiao Song, Yanbo Fan, Aoyang Zhou, Baoyuan Wu, Yiming Li, Zhifeng Li, Kun He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial region-based sampler (ARS) to SAT (ablation) | SAT · CIFAR-10 | PGD robust accuracy: 47.74 -> 52.59 | Improved |
| add ARS to SAT (ablation) | SAT · CIFAR-10 | clean accuracy: 86.92 -> 87.67 | Improved |
| add ARS to SAT (ablation) | SAT · CIFAR-10 | CW-inf robust accuracy: 48.64 -> 50.87 | Improved |
| add distance-aware label smoothing (DLS) on top of SAT+ARS (ablation) | SAT+ARS · CIFAR-10 | PGD robust accuracy: 52.59 -> 70.57 | Improved |
| add DLS on top of SAT+ARS (ablation) | SAT+ARS · CIFAR-10 | clean accuracy: 87.67 -> 93.17 | Improved |
| add DLS on top of SAT+ARS (ablation) | SAT+ARS · CIFAR-10 | CW-inf robust accuracy: 50.87 -> 65.34 | Improved |
| raise beta_min from 0.1 to 0.3 (ablation) | RAT · CIFAR-10 | PGD robust accuracy: 70.57 -> 64.87 | Hurt |
| raise beta_min from 0.1 to 0.5 (ablation) | RAT · CIFAR-10 | PGD robust accuracy: 70.57 -> 62.86 | Hurt |
| lower beta_max from 1.0 to 0.8 (ablation) | RAT · CIFAR-10 | PGD robust accuracy: 70.57 -> 66.44 | Hurt |
| lower beta_max from 1.0 to 0.6 (ablation) | RAT · CIFAR-10 | PGD robust accuracy: 70.57 -> 62.93 | Hurt |
| replace SAT with RAT | SAT · CIFAR-10 | PGD robust accuracy: 47.74 -> 70.57 | Improved |
| replace SAT with RAT | SAT · CIFAR-100 | PGD robust accuracy: 24.34 -> 37.03 | Improved |
Closest papers by meaning
- Variational Adversarial Defense: A Bayes Perspective for Adversarial Training (TPAMI 2023 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- Taxonomy Driven Fast Adversarial Training (AAAI 2024 · similarity 0.88)
- Enhancing the Adversarial Robustness via Manifold Projection (AAAI 2025 · similarity 0.88)
- A Closer Look at Curriculum Adversarial Training: From an Online Perspective (AAAI 2024 · similarity 0.88)
- Clarifying the Behavior and the Difficulty of Adversarial Training (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
