Research map / Adversarial robustness and security / Certified and empirical robustness
Understanding and Mitigating the Tradeoff between Robustness and Accuracy
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 RST to PG-AT using unlabeled data | PG-AT · CIFAR-10, ℓ∞, ε=8/255 | robust test accuracy: 45.8% -> 58.5% | Helped |
| Apply RST to PG-AT using unlabeled data | PG-AT · CIFAR-10, ℓ∞, ε=8/255 | standard test accuracy: 87.3% -> 91.8% | Helped |
| Apply RST to TRADES using unlabeled data | TRADES · CIFAR-10, ℓ∞, ε=8/255 | robust test accuracy: 55.4% -> 63.1% | Helped |
| Apply RST to TRADES using unlabeled data | TRADES · CIFAR-10, ℓ∞, ε=8/255 | standard test accuracy: 84.0% -> 89.7% | Helped |
| Apply RST to Worst-of-10 using unlabeled data | Worst-of-10 · CIFAR-10, rotations and translations | robust test accuracy: 73.9% -> 75.1% | Helped |
| Apply RST to Worst-of-10 using unlabeled data | Worst-of-10 · CIFAR-10, rotations and translations | standard test accuracy: 95.0% -> 95.8% | Helped |
| Apply RST to random augmentation using unlabeled data | Random · CIFAR-10, rotations and translations | robust test accuracy: 67.7% -> 70.9% | Helped |
| Apply RST to random augmentation using unlabeled data | Random · CIFAR-10, rotations and translations | standard test accuracy: 95.1% -> 95.8% | Helped |
| Use only standard loss on unlabeled pseudo-labeled data, omitting robust loss (ablation) | Robust self-training · CIFAR-10, ℓ∞, ε=8/255 | robust test accuracy: 0.3% | Hurt |
| Use only robust self-consistency loss on unlabeled data, omitting standard pseudo-label loss (ablation) | Robust self-training · CIFAR-10, ℓ∞, ε=8/255 | standard test accuracy: 83.2% | Hurt |
Closest papers by meaning
- Precise Accuracy / Robustness Tradeoffs in Regression: Case of General Norms (ICML 2024 · similarity 0.91)
- Robustness May Be at Odds with Accuracy (ICLR 2019 · similarity 0.90)
- Theoretically Principled Trade-off between Robustness and Accuracy (ICML 2019 · similarity 0.90)
- Data Augmentation Can Improve Robustness (NeurIPS 2021 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- On the Duality Between Sharpness-Aware Minimization and Adversarial Training (ICML 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.