Research map / Adversarial robustness and security / Certified and empirical robustness
Towards Evaluating the Robustness of Neural Networks Learned by Transduction
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 |
|---|---|---|---|
| Evaluate RMC with GMSA-AVG rather than the defense-unaware AutoAttack | RMC with a standard base model · MNIST | robustness: 97.70% to 0.50% | Hurt |
| Evaluate RMC with GMSA-AVG rather than the defense-unaware AutoAttack | RMC with a standard base model · CIFAR-10 | robustness: 94.20% to 8.00% | Hurt |
| Apply RMC on an adversarially trained model instead of using adversarial training alone | RMC · MNIST and CIFAR-10 | robustness: 96.10% to 58.80%; 71.70% to 39.60% | Hurt |
| Use GMSA to evaluate DANN instead of AutoAttack | DANN · MNIST and CIFAR-10 | robustness: 97.59 to 6.17; 66.61 to 5.53 | Hurt |
| Adversarially retrain at test time using fresh private randomness | TADV · MNIST | robustness: 86.61% to 94.27% | Helped |
| Adversarially retrain at test time using fresh private randomness | TADV · CIFAR-10 | robustness: 45.29% to 54.12% | Helped |
Closest papers by meaning
- Two Heads are Actually Better than One: Towards Better Adversarial Robustness via Transduction and Rejection (ICML 2024 · similarity 0.91)
- Efficient Robustness Evaluation via Constraint Relaxation (AAAI 2025 · similarity 0.87)
- Rethinking the Adversarial Robustness of Multi-Exit Neural Networks in an Attack-Defense Game (CVPR 2025 · similarity 0.86)
- Certified Causal Defense with Generalizable Robustness (AAAI 2025 · similarity 0.86)
- Model Inversion Robustness: Can Transfer Learning Help? (CVPR 2024 · similarity 0.85)
- On the Scalability of Certified Adversarial Robustness with Generated Data (NeurIPS 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.