Research map / Adversarial robustness and security / Certified and empirical robustness
To be Robust or to be Fair: Towards Fairness in 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 PGD adversarial training instead of natural training | CIFAR-10 classifier · CIFAR-10 | WD: 5.5 -> 18.6 | Hurt |
| Apply PGD adversarial training instead of natural training | CIFAR-10 classifier · CIFAR-10 | NSD: 0.47 -> 0.57 | Hurt |
| Use FRL with reweighting and remargining instead of PGD adversarial training | CIFAR-10 classifier · CIFAR-10 | WD: 18.6 -> 9.7 | Helped |
| Use FRL with reweighting and remargining instead of PGD adversarial training | CIFAR-10 classifier · CIFAR-10 | NSD: 0.57 -> 0.43 | Helped |
| Add remargining to FRL reweighting (ablation) | FRL (RW) · CIFAR-10 | WD: 10.2 -> 9.7 | Helped |
| Add reweighting to FRL remargining (ablation) | FRL (RM) · CIFAR-10 | WD: 9.3 -> 9.7 | Hurt |
| Use remargining rather than reweighting (ablation) | FRL · CIFAR-10 | WCBE during training: 45 -> 43 | Helped |
| Combine reweighting and remargining (ablation) | FRL · CIFAR-10 | WCBE during training: 48 -> 39 | Helped |
| Combine reweighting and remargining (ablation) | FRL · CIFAR-10 | WCBE test: 56 -> 45 | Helped |
| Minimize class-wise deviation instead of worst-class discrepancy (ablation) | CIFAR-10 adversarial training · CIFAR-10 | average robust error: 56% compared to FRLs | Hurt |
| Use FRL instead of PGD adversarial training | CIFAR-10 classifier · CIFAR-10 | average robust error: reduces avg. robust error by 2-3% | Helped |
| Use FRL instead of PGD adversarial training | CIFAR-10 classifier · CIFAR-10 | average standard error: increases the avg. standard error by 0.5-1.5% | Hurt |
Closest papers by meaning
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.94)
- DAFA: Distance-Aware Fair Adversarial Training (ICLR 2024 · similarity 0.92)
- Towards Fairness-Aware Adversarial Learning (CVPR 2024 · similarity 0.92)
- Improving Adversarial Robust Fairness via Anti-Bias Soft Label Distillation (NeurIPS 2024 · similarity 0.91)
- Can Private Machine Learning Be Fair? (AAAI 2025 · similarity 0.90)
- On the Fairness ROAD: Robust Optimization for Adversarial Debiasing (ICLR 2024 · similarity 0.90)
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.