Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Neural Pruning with Latent Vulnerability Suppression
Read the paperCheck 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 |
|---|---|---|---|
| Add vulnerability suppression loss to adversarial training (ablation) | AT · MNIST | white-box adversarial accuracy: 88.03{ $\pm$0.7} -> 90.36{ $\pm$0.8} | Helped |
| Add vulnerability suppression loss to adversarial training (ablation) | AT · CIFAR-10 | white-box adversarial accuracy: 49.85{ $\pm$0.9} -> 51.52{ $\pm$0.4} | Helped |
| Add vulnerability suppression loss to adversarial training (ablation) | AT · CIFAR-100 | white-box adversarial accuracy: 19.07{ $\pm$0.8} -> 20.06{ $\pm$0.9} | Helped |
| Add adversarial neural pruning without VS loss (ablation) | AT · MNIST | white-box adversarial accuracy: 88.03{ $\pm$0.7} -> 90.12{ $\pm$0.5} | Helped |
| Add adversarial neural pruning without VS loss (ablation) | AT · CIFAR-10 | white-box adversarial accuracy: 49.85{ $\pm$0.9} -> 55.63{ $\pm$0.9} | Helped |
| Add adversarial neural pruning without VS loss (ablation) | AT · CIFAR-100 | white-box adversarial accuracy: 19.07{ $\pm$0.8} -> 22.20{ $\pm$1.1} | Helped |
| Combine pruning with vulnerability suppression loss (ablation) | ANP · MNIST | white-box adversarial accuracy: 90.12{ $\pm$0.5} -> 91.31{ $\pm$0.9} | Helped |
| Combine pruning with vulnerability suppression loss (ablation) | ANP · CIFAR-10 | white-box adversarial accuracy: 55.63{ $\pm$0.9} -> 56.21{ $\pm$0.1} | Helped |
| Combine pruning with vulnerability suppression loss (ablation) | ANP · CIFAR-100 | white-box adversarial accuracy: 22.20{ $\pm$1.1} -> 22.35{ $\pm$0.6} | Helped |
| Increase sparsity beyond approximately 80% (ablation) | ANP-VS · CIFAR-10 and CIFAR-100 | robust and clean accuracy | Hurt |
| Prune first, freeze the mask, then perform adversarial training (ablation) | PAT · CIFAR-10 and CIFAR-100 | clean accuracy | Hurt |
Closest papers by meaning
- Masking Adversarial Damage: Finding Adversarial Saliency for Robust and Sparse Network (CVPR 2022 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models (ICML 2025 · similarity 0.87)
- TAET: Two-Stage Adversarial Equalization Training on Long-Tailed Distributions (CVPR 2025 · similarity 0.86)
- Robust Overfitting Does Matter: Test-Time Adversarial Purification With FGSM (CVPR 2024 · similarity 0.86)
- Layer-Aware Analysis of Catastrophic Overfitting: Revealing the Pseudo-Robust Shortcut Dependency (ICML 2024 · similarity 0.86)
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.