Research map / Adversarial robustness and security / Certified and empirical robustness
Combining Adversaries with Anti-adversaries in Training
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 |
|---|---|---|---|
| Combine adversarial and anti-adversarial perturbations (ablation) | CAAT · CIFAR10 | Avg. Rob. (%): 57.6 -> 28.7 | Helped |
| Use different bounds while keeping the same perturbation direction (ablation) | CAAT · CIFAR10 | Avg. Rob. (%): 57.6 -> 51.9 | Helped |
| Combine adversarial and anti-adversarial directions with the same bound (ablation) | CAAT · CIFAR10 | Avg. Rob. (%): 51.9 -> 28.7 | Helped |
| Use different bounds in addition to combining perturbation directions (ablation) | CAAT · CIFAR10 | Avg. Nat. (%): 14.9 -> 13.9 | Helped |
| Use different bounds in addition to combining perturbation directions (ablation) | CAAT · CIFAR10 | Avg. Bdy. (%): 13.8 -> 15.4 | Hurt |
| Use different bounds in addition to combining perturbation directions (ablation) | CAAT · CIFAR10 | Avg. Rob. (%): 28.7 -> 29.3 | Hurt |
| Use anti-adversarial perturbations for noisy samples and adversarial perturbations for clean samples | CAAT · CIFAR10 with 20% pair-flip noise | Avg. Rob. (%): 37.4 -> 28.5 | Helped |
| Use sample-wise Grad-Based bounds instead of ReMargin bounds | CAAT · CIFAR10 with 20% pair-flip noise | Avg. Rob. (%): 29.4 -> 28.5 | Helped |
| Use adversarial and anti-adversarial training with learned sample weights | CAAT · CIFAR10 with imbalance factor 10 | Avg. Rob. (%): 47.4 -> 29.9 | Helped |
| Use ReMargin bounds instead of Grad-Based bounds | CAAT · CIFAR10 with imbalance factor 10 | Avg. Bdy. (%): 14.2 -> 13.7 | Helped |
| Use ReMargin bounds instead of Grad-Based bounds | CAAT · CIFAR10 | Worst Rob. (%): 48.1 -> 44.4 | Helped |
Closest papers by meaning
- Adversarial Training and Robustness for Multiple Perturbations (NeurIPS 2019 · similarity 0.89)
- A Simple Class-Agnostic Approach to Enhance Fair Adversarial Training (NeurIPS 2026 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.87)
- Clarifying the Behavior and the Difficulty of Adversarial Training (AAAI 2024 · similarity 0.87)
- Improving Adversarial Robust Fairness via Anti-Bias Soft Label Distillation (NeurIPS 2024 · similarity 0.87)
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.