Research map / Adversarial robustness and security / Certified and empirical robustness
Are Adversarial Examples Created Equal? A Learnable Weighted Minimax Risk for Robustness under Non-uniform Attacks
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 adaptive margin-aware weighting to TRADES | TRADES · CIFAR10, non-uniform attack (α_test=2.0) | A_tr (%): 17.14 -> 19.79 | Helped |
| Add adaptive margin-aware weighting to PGD adversarial training | PGD · CIFAR10, non-uniform attack (α_test=2.0) | A_tr (%): 9.72 -> 10.81 | Helped |
| Add adaptive margin-aware weighting to TRADES | TRADES · MNIST, non-uniform attack (α_test=2.0) | A_tr (%): 51.52 -> 65.61 | Helped |
| Add adaptive margin-aware weighting to TRADES | TRADES · Tiny ImageNet, non-uniform attack (α_test=2.0) | A_tr (%): 12.58 -> 15.01 | Helped |
| Add adaptive margin-aware weighting to TRADES | TRADES · CIFAR10, uniform attack | robust accuracy (%): 53.38 -> 54.10 | Helped |
| Increase α_train from 0.1 to 2.0 (ablation) | TRADES+ours · CIFAR10, non-uniform attack (α_test=2.0) | A_tr (%): 17.59 -> 19.79 | Helped |
| Increase α_train from 0.1 to 0.1 (ablation) | PGD+ours · CIFAR10, non-uniform attack (α_test=2.0) | A_tr (%): 10.01 -> 10.81 | Helped |
| Use adaptive margin-aware weighting with α_train=0.5 (ablation) | TRADES · MNIST, uniform attack | robust accuracy (%): 95.59 -> 94.69 | Hurt |
| Vary α_train across tested values (ablation) | TRADES+ours · CIFAR10, margin-loss PGD evaluation | robust accuracy (%): 49.50 -> 50.34; 49.50 -> 48.95 | Mixed |
| Add margin-aware weighting to PGD adversarial training (ablation) | PGD · CIFAR10, uniform attack | robust accuracy (%): 49.29 -> 49.08; 49.29 -> 49.53 | Mixed |
Closest papers by meaning
- Are adversarial examples inevitable? (ICLR 2019 · similarity 0.93)
- Certified Adversarial Robustness with Additive Noise (NeurIPS 2019 · similarity 0.92)
- Adversarial Risk and the Dangers of Evaluating Against Weak Attacks (ICML 2018 · similarity 0.91)
- On Adversarial Training without Perturbing all Examples (ICLR 2024 · similarity 0.91)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.90)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.89)
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.