Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Vertex Mixup: Toward Better Adversarially Robust Generalization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace PGD training with AVmixup | CIFAR10 model · CIFAR10 | PGD20 accuracy: 46.34 -> 58.23 | Helped |
| Use AVmixup instead of PGD training | CIFAR10 model · CIFAR10 | Clean accuracy: 86.88 -> 93.24 | Helped |
| Use AVmixup rather than TRADES (1 / λ = 6) | CIFAR10 model · CIFAR10 | PGD20 accuracy: 56.61 -> 58.27 | Helped |
| Use AVmixup rather than PGD training | CIFAR100 model · CIFAR100 | PGD20 accuracy: 25.17 -> 38.49 | Helped |
| Use AVmixup rather than PGD training | SVHN model · SVHN | PGD20 accuracy: 58.22 -> 61.90 | Helped |
| Use AVmixup rather than PGD training | Tiny ImageNet model · Tiny ImageNet | PGD20 accuracy: 13.14 -> 20.31 | Helped |
| Add AVmixup to Feature Scatter training | Feature Scatter · CIFAR10 | PGD20 accuracy: 69.74 -> 82.31 | Helped |
| Add AVmixup to Feature Scatter training | Feature Scatter · CIFAR100 | CW20 accuracy: 27.35 -> 33.20 | Helped |
| Use AVmixup with PGD-based training | CIFAR100 model · CIFAR100 | CW20 accuracy: 22.98 -> 23.46; LS0.8: 26.15 | Mixed |
| Increase γ from 1.0 to 2.0 (ablation) | Feature Scatter + AVmixup with λ1=0.5, λ2=0.7 · CIFAR10 | PGD20 accuracy: 82.31 -> 52.87 | Hurt |
| Set λ1=1.0 and λ2=0.5 rather than λ1=0.5 and λ2=0.7 (ablation) | Feature Scatter + AVmixup, γ=1.0 · CIFAR10 | PGD20 accuracy: 82.31 -> 53.42 | Hurt |
| Set λ1=0.5 and λ2=0.3 rather than λ1=0.5 and λ2=0.7 (ablation) | Feature Scatter + AVmixup, γ=1.0 · CIFAR10 | PGD20 accuracy: 82.31 -> 57.67 | Hurt |
Closest papers by meaning
- How Does Mixup Help With Robustness and Generalization? (ICLR 2021 · similarity 0.91)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- Adversarial AutoMixup (ICLR 2024 · similarity 0.89)
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 2024 · similarity 0.88)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.88)
- Benign Overfitting in Adversarial Training of Neural Networks (ICML 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.