Research map / Adversarial robustness and security / Certified and empirical robustness
Efficient Certification of Spatial Robustness
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 |
|---|---|---|---|
| Add flow constraints with γ=0.001 (ablation) | DeepPoly on ConvSmall PGD · MNIST, δ=0.4 | certification rate (%): 6 -> 77 | Helped |
| Add flow constraints with γ=0.001 (ablation) | MILP on ConvSmall DiffAI · CIFAR-10, δ=0.4 | certification rate (%): 38 -> 69 | Helped |
| Add flow constraints with γ=0.1 (ablation) | DeepPoly on CIFAR-10 ResNet · CIFAR-10, δ=0.4 | certification rate (%): 72% -> 76% | Helped |
| Add flow constraints with γ=0.1 (ablation) | MILP on CIFAR-10 ResNet · CIFAR-10, δ=0.4 | certification rate (%): 87% -> 89% | Helped |
| Use T1 rather than T∞ displacement norm (ablation) | DeepPoly on ConvSmall DiffAI · MNIST, δ=0.5, γ=∞ | certification rate (%): 40 -> 75 | Helped |
| Use T1 rather than T∞ displacement norm (ablation) | MILP on ConvSmall DiffAI · MNIST, δ=0.5, γ=∞ | certification rate (%): 50 -> 79 | Helped |
Closest papers by meaning
- Certifying Geometric Robustness of Neural Networks (NeurIPS 2019 · similarity 0.88)
- Exploring the Landscape of Spatial Robustness (ICML 2019 · similarity 0.87)
- Can we have it all? On the Trade-off between Spatial and Adversarial Robustness of Neural Networks (NeurIPS 2021 · similarity 0.87)
- GaLileo: General Linear Relaxation Framework for Tightening Robustness Certification of Transformers (AAAI 2024 · similarity 0.86)
- Stop Walking in Circles! Bailing Out Early in Projected Gradient Descent (CVPR 2025 · similarity 0.84)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.84)
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.