Research map / Adversarial robustness and security / Certified and empirical robustness
A Unified Wasserstein Distributional Robustness Framework for 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 |
|---|---|---|---|
| Add Wasserstein distributional robustness to PGD-AT | PGD-AT · MNIST | AA accuracy: 88.9 -> 90.0 | Helped |
| Add Wasserstein distributional robustness to PGD-AT | PGD-AT · CIFAR10 | AA accuracy: 42.5 -> 44.8 | Helped |
| Add Wasserstein distributional robustness to PGD-AT | PGD-AT · CIFAR100 | AA accuracy: 39.3 -> 41.9 | Helped |
| Add Wasserstein distributional robustness to TRADES | TRADES · MNIST | AA accuracy: 90.9 -> 92.2 | Helped |
| Add Wasserstein distributional robustness to TRADES | TRADES · CIFAR10 | AA accuracy: 49.1 -> 49.9 | Helped |
| Add Wasserstein distributional robustness to TRADES | TRADES · CIFAR100 | AA accuracy: 46.7 -> 47.8 | Helped |
| Add Wasserstein distributional robustness to MART | MART · MNIST | AA accuracy: 90.6 -> 92.3 | Helped |
| Add Wasserstein distributional robustness to MART | MART · CIFAR10 | AA accuracy: 48.2 -> 49.1 | Helped |
| Add Wasserstein distributional robustness to MART | MART · CIFAR100 | AA accuracy: 44.8 -> 48.5 | Helped |
| Increase hard-ball PGD-AT perturbation limit from ε=0.031 to ε=0.034 (ablation) | PGD-AT · CIFAR10 | Avg: 0.75% | Helped |
| Increase hard-ball PGD-AT perturbation limit from ε=0.031 to ε=0.037 (ablation) | PGD-AT · CIFAR10 | Avg: -0.07% | Hurt |
| Replace hard-ball projection with adaptive soft-ball projection (ablation) | PGD-AT · CIFAR10 | Avg: 3.08 | Helped |
Closest papers by meaning
- Wasserstein distributional robustness of neural networks (NeurIPS 2023 · similarity 0.94)
- Certifying Some Distributional Robustness with Principled Adversarial Training (ICLR 2018 · similarity 0.92)
- Provable Robust Overfitting Mitigation in Wasserstein Distributionally Robust Optimization (ICLR 2025 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- Universal generalization guarantees for Wasserstein distributionally robust models (ICLR 2025 · similarity 0.89)
- DRF: Improving Certified Robustness via Distributional Robustness Framework (AAAI 2024 · similarity 0.88)
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.