Research map / Adversarial robustness and security / Certified and empirical robustness
Geometry-aware Instance-reweighted Adversarial 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 |
|---|---|---|---|
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, best checkpoint | PGD-20 accuracy: 51.96 ± 0.21 -> 57.81 ± 0.54; +5.85 | Helped |
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, best checkpoint | PGD+ accuracy: 51.28 ± 0.23 -> 55.61 ± 0.61; +4.33 | Helped |
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, best checkpoint | natural accuracy: 86.92 ± 0.24 -> 85.75 ± 0.23; -1.17 | Hurt |
| Add geometry-aware instance reweighting to FAT | WRN-32-10 · CIFAR-10 test, best checkpoint | PGD-20 accuracy: 51.24 ± 0.14 -> 56.21 ± 0.52 | Helped |
| Add geometry-aware instance reweighting to FAT | WRN-32-10 · CIFAR-10 test, best checkpoint | natural accuracy: 89.16 ± 0.15 -> 88.59 ± 0.12 | Hurt |
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, last checkpoint | PGD-20 accuracy: 46.73 ± 0.08 -> 53.76 ± 0.49; +7.03 | Helped |
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, last checkpoint | PGD+ accuracy: 46.08 ± 0.07 -> 50.32 ± 0.48; +4.24 | Helped |
| Apply geometry-aware instance reweighting to AT | WRN-32-10 · CIFAR-10 test, last checkpoint | natural accuracy: 86.62 ± 0.22 -> 85.49 ± 0.25; -1.13 | Hurt |
| Use geometry-based decreasing weights instead of equal AT weights (ablation) | ResNet-18 · CIFAR-10 | robust test accuracy | Helped |
| Use the tanh-type decreasing weight function rather than alternative tested weight functions (ablation) | GAIRAT · CIFAR-10 | robust test accuracy | Helped |
| Decrease lambda to emphasize attackable examples more (ablation) | GAIRAT · CIFAR-10 | robust test accuracy | Helped |
| Shorten the burn-in period before geometry-based weighting (ablation) | GAIRAT · CIFAR-10 | robust test accuracy | Helped |
Closest papers by meaning
- Doubly Robust Instance-Reweighted Adversarial Training (ICLR 2024 · similarity 0.91)
- Improving Accuracy-robustness Trade-off via Pixel Reweighted Adversarial Training (ICML 2024 · similarity 0.90)
- Bag of Tricks for Adversarial Training (ICLR 2021 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- A Simple Class-Agnostic Approach to Enhance Fair Adversarial Training (NeurIPS 2026 · similarity 0.89)
- Robustness via Curvature Regularization, and Vice Versa (CVPR 2019 · 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.