Research map / Adversarial robustness and security / Certified and empirical robustness
You Only Propagate Once: Accelerating Adversarial Training via Maximal Principle
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 |
|---|---|---|---|
| Replace PGD-40 adversarial training with YOPO-5-10 | MNIST adversarial training · MNIST | PGD-40 Attack accuracy: 96.56% -> 96.27% | No clear effect |
| Replace PGD-40 adversarial training with YOPO-5-10 | MNIST adversarial training · MNIST | CW Attack accuracy: 93.52% -> 93.56% | No clear effect |
| Replace PGD-40 adversarial training with YOPO-5-10 | MNIST adversarial training · MNIST | training time: 250 seconds; 1250s | Helped |
| Replace PGD-10 adversarial training with YOPO-5-3 | PreAct-Res18 · CIFAR10 | PGD-20 Attack accuracy: 41.61% -> 44.72% | Helped |
| Replace PGD-10 adversarial training with YOPO-5-3 | PreAct-Res18 · CIFAR10 | CW Attack accuracy: 58.88% -> 59.77% | Helped |
| Replace PGD-10 adversarial training with YOPO-5-3 | Wide ResNet34 · CIFAR10 | PGD-20 Attack accuracy: 47.04% -> 47.98% | Helped |
| Replace PGD-10 adversarial training with YOPO-5-3 | Wide ResNet34 · CIFAR10 | Training Time: 2713 -> 476 | Helped |
| Replace Free-8 adversarial training with YOPO-5-3 | Wide ResNet34 · CIFAR10 | PGD-20 Attack accuracy: 47.00% -> 47.98% | Helped |
| Replace TRADES-10 with TRADES-YOPO-3-4 | PreAct-Res18 with TRADES objective · CIFAR10 | PGD-20 Attack accuracy: 44.50% -> 46.13% | Helped |
| Replace TRADES-10 with TRADES-YOPO-3-4 | PreAct-Res18 with TRADES objective · CIFAR10 | Training Time: 633 -> 259 | Helped |
| Vary YOPO configuration from 5-3 to 3-5 (ablation) | YOPO adversarial training on Wide ResNet34 · CIFAR10 | PGD-20 Attack accuracy: 47.98% -> 43.04% | Hurt |
| Vary YOPO configuration from 5-3 to 3-5 (ablation) | YOPO adversarial training on Wide ResNet34 · CIFAR10 | Training Time: 476 -> 299 | Helped |
Closest papers by meaning
- Clarifying the Behavior and the Difficulty of Adversarial Training (AAAI 2024 · similarity 0.89)
- Taxonomy Driven Fast Adversarial Training (AAAI 2024 · similarity 0.88)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- A Closer Look at Curriculum Adversarial Training: From an Online Perspective (AAAI 2024 · similarity 0.87)
- Adversarial Training Should Be Cast as a Non-Zero-Sum Game (ICLR 2024 · similarity 0.87)
- Efficient Adversarial Training in LLMs with Continuous Attacks (NeurIPS 2024 · similarity 0.87)
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.