Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Models are less Over-Confident
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 adversarial training | CIFAR10 model pairs · CIFAR10 clean and PGD | ECE: 0.6736 $\pm$ 0.1208 -> 0.1894 $\pm$0.1531; 0.6809 $\pm$ 0.1061 -> 0.2688 $\pm$ 0.1733 | Helped |
| Apply adversarial training | CIFAR100 model pairs · CIFAR100 clean | ECE: 0.3077 $\pm$ 0.1257 -> 0.2962 $\pm$0.1722 | Helped |
| Apply adversarial training | CIFAR100 model pairs · CIFAR100 PGD | ECE: 0.2159 $\pm$ 0.0738 -> 0.2307 $\pm$ 0.1494 | Hurt |
| Use improved downsampling (ablation) | PRN-18 · CIFAR10 PGD | AUC: 0.0942 -> 0.9832 | Helped |
| Use improved downsampling (ablation) | PRN-18 · CIFAR10 Squares | AUC: 0.8347 -> 0.9923 | Helped |
| Use improved activation (ablation) | WRN-28-10 · CIFAR10 PGD | AUC: 0.2781 -> 0.9755 | Helped |
| Use improved activation (ablation) | WRN-28-10 · CIFAR10 Squares | AUC: 0.9076 -> 0.9877 | Helped |
| Use improved downsampling or activation (ablation) | Robust CIFAR10 models · CIFAR10-C | calibration | No clear effect |
| Apply adversarial training | CIFAR models · Squares black-box attack | accuracy | Mixed |
| Apply adversarial training | CIFAR models · PGD white-box attack | accuracy | Helped |
| Use a larger adversarially trained model | ImageNet models · ImageNet PGD | accuracy: 31.92 -> 44.75 | Helped |
Closest papers by meaning
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.94)
- Bag of Tricks for Adversarial Training (ICLR 2021 · similarity 0.92)
- On the Vulnerability of Adversarially Trained Models Against Two-faced Attacks (ICLR 2024 · similarity 0.92)
- Benchmarking Adversarial Robustness on Image Classification (CVPR 2020 · similarity 0.92)
- WAT: Improve the Worst-Class Robustness in Adversarial Training (AAAI 2023 · similarity 0.92)
- A Combination of Noise and Bilateral Filters Achieve Supralinear and Scalable Adversarial Robustness in CNNs (CVPR 2026 · similarity 0.92)
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.