Research map / Adversarial robustness and security / Certified and empirical robustness
Almost Tight L0-norm Certified Robustness of Top-k Predictions against Adversarial Perturbations
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 |
|---|---|---|---|
| Extend randomized ablation to certify top-3 predictions instead of transforming Gaussian top-k ℓ2 certificates to ℓ0 | Randomized ablation · CIFAR10, 1 perturbed pixel | certified top-3 accuracy: 0.244 -> 0.886 | Helped |
| Extend randomized ablation to certify top-3 predictions instead of transforming Gaussian top-k ℓ2 certificates to ℓ0 | Randomized ablation · ImageNet, 1 perturbed pixel | certified top-3 accuracy: 0.326 -> 0.740 | Helped |
| Estimate label probability bounds with SimuEM | Randomized ablation, top-1 · CIFAR10 | certified top-1 accuracy: 0.704 -> 0.746 | Helped |
| Estimate label probability bounds with SimuEM | Randomized ablation, top-1 · ImageNet | certified top-1 accuracy: 0.602 -> 0.634 | Helped |
| Increase k (ablation) | Our method · CIFAR10 and ImageNet | certified top-k accuracy: 63.6% and 78.0% | Helped |
| Increase the number of retained features e (ablation) | Our method · CIFAR10 and ImageNet | certified accuracy | Mixed |
| Increase the number of Monte Carlo samples n (ablation) | Our method · CIFAR10 and ImageNet | certified top-3 accuracy | Helped |
| Decrease the confidence level 1−α (ablation) | Our method · CIFAR10 and ImageNet | certified top-3 accuracy | Helped |
| Vary α among high-confidence settings (ablation) | Our method · CIFAR10 and ImageNet | certified top-3 accuracy | No clear effect |
Closest papers by meaning
- Certified Robustness for Top-k Predictions against Adversarial Perturbations via Randomized Smoothing (ICLR 2020 · similarity 0.95)
- Relaxing Local Robustness (NeurIPS 2021 · similarity 0.90)
- Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing (ICML 2025 · similarity 0.88)
- $\sigma$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples (ICLR 2025 · similarity 0.88)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.88)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · 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.