Research map / Adversarial robustness and security / Certified and empirical robustness
Globally-Robust Neural Networks
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 |
|---|---|---|---|
| Train with GloRo Nets instead of prior deterministic certification methods | certified robust classification · MNIST, ε=1.58 | verified-robust accuracy: approximately 15 percentage points | Helped |
| Train with GloRo Nets instead of BCP | certified robust classification · CIFAR-10 | verified-robust accuracy: approximately 7 percentage points | Helped |
| Train with GloRo Nets instead of BCP | certified robust classification · Tiny-Imagenet | verified-robust accuracy: approximately 2 percentage points | Helped |
| Use MinMax activations instead of ReLU activations (ablation) | GloRo Nets · benchmarks evaluated | verified-robust accuracy | Helped |
| Use the GloRo-TRADES loss instead of cross-entropy training (ablation) | GloRo Nets · benchmarks evaluated | verified-robust accuracy | Helped |
| Use cross-entropy for both terms instead of KL divergence in the second GloRo-TRADES term (ablation) | GloRo-TRADES · benchmarks evaluated | verified-robust accuracy | No clear effect |
| Replace the full Tiny-Imagenet architecture with one fewer dense layer and one fewer convolutional layer (ablation) | GloRo Net on Tiny-Imagenet · Tiny-Imagenet | verified-robust accuracy | No clear effect |
| Use the lower bound instead of the upper bound for the global Lipschitz constant (ablation) | GloRo Net · CIFAR-10 | verified-robust accuracy: 58% to 64% | Helped |
| Use a global Lipschitz bound rather than local-bound certification | GloRo Nets compared with BCP · CIFAR-10 | certification speed: over 14× faster | Helped |
Closest papers by meaning
- A Recipe for Improved Certifiable Robustness (ICLR 2024 · similarity 0.92)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.90)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.90)
- Achieving Domain-Independent Certified Robustness via Knowledge Continuity (NeurIPS 2024 · similarity 0.89)
- Efficient Robust Conformal Prediction via Lipschitz-Bounded Networks (ICML 2025 · similarity 0.89)
- 1-Lipschitz Layers Compared: Memory Speed and Certifiable Robustness (CVPR 2024 · similarity 0.88)
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.