Research map / Adversarial robustness and security / Certified and empirical robustness
Mixup Inference: Better Exploiting Mixup to Defend Adversarial Attacks
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 MI-Combined during inference | Mixup-trained model · CIFAR-10, oblivious untargeted PGD_10 | accuracy: 3.6 -> 33.7 | Helped |
| Apply MI-OL during inference | Interpolated AT · CIFAR-10, oblivious untargeted PGD_10 | accuracy: 46.7 -> 64.5 | Helped |
| Apply MI-OL during inference | Interpolated AT · CIFAR-10, oblivious targeted PGD_200 | accuracy: 61.9 -> 74.5 | Helped |
| Apply MI-Combined during inference | Mixup-trained model · CIFAR-100, oblivious untargeted PGD_10 | accuracy: 5.5 -> 14.8 | Helped |
| Apply MI-OL during inference | Interpolated AT · CIFAR-100, oblivious untargeted PGD_10 | accuracy: 26.6 -> 43.8 | Helped |
| Apply MI-OL during inference | Interpolated AT · CIFAR-100, oblivious targeted PGD_200 | accuracy: 49.8 -> 56.5 | Helped |
| Apply MI-OL during inference (ablation) | ERM-trained model · CIFAR-10, oblivious untargeted PGD_10 | accuracy | Helped |
| Replace mixup-based training with ERM training (ablation) | MI-OL defense · CIFAR-10, oblivious untargeted PGD_10 | accuracy: 26.1 -> 7.4 | Hurt |
| Replace interpolated AT with AT training (ablation) | MI-OL defense · CIFAR-10, oblivious untargeted PGD_10 | accuracy: 64.5 -> 56.2 | Hurt |
| Use MI-Combined instead of MI-OL (ablation) | Mixup-trained model · CIFAR-10, oblivious targeted PGD_200 | accuracy: 50.8 -> 49.4 | Hurt |
Closest papers by meaning
- mixup: Beyond Empirical Risk Minimization (ICLR 2018 · similarity 0.91)
- MixAT: Combining Continuous and Discrete Adversarial Training for LLMs (NeurIPS 2025 · similarity 0.88)
- Pushing Boundaries: Mixup's Influence on Neural Collapse (ICLR 2024 · similarity 0.87)
- Improving Transferable Targeted Attacks with Feature Tuning Mixup (CVPR 2025 · similarity 0.87)
- Adversarial Mixup Unlearning (ICLR 2025 · similarity 0.86)
- Adversarial AutoMixup (ICLR 2024 · similarity 0.85)
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.