Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Robustness Across Representation Spaces
Check 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 |
|---|---|---|---|
| use multiplicative weights (h=3) instead of the strongest heuristic baseline | CIFAR-10 multi-attack training · CIFAR-10 | Min. Accuracy: 29.99 ± 1.49 -> 39.13 ± 1.74 | Helped |
| use multiplicative weights (h=3) instead of the strongest heuristic baseline | CIFAR-10 multi-attack training · CIFAR-10 | Union Attack: 31.95 ± 4.55 -> 36.18 ± 1.89 | Helped |
| use multiplicative weights (h=3) instead of the strongest heuristic baseline | MNIST multi-attack training · MNIST | Min. Accuracy: 22.13 ± 17.17 -> 57.57 ± 8.46 | Helped |
| use multiplicative weights (h=3) instead of the strongest heuristic baseline | MNIST multi-attack training · MNIST | Union Attack: 12.32 ± 9.80 -> 35.87 ± 6.69 | Helped |
| average the last three model parameters rather than use the last time step (ablation) | Multiplicative Weights · CIFAR-10 | Min. Accuracy: 35.70 ± 6.36 -> 39.13 ± 1.74 | Helped |
| average the last three model parameters rather than use the last time step (ablation) | Multiplicative Weights · MNIST | Union Attack: 35.30 ± 4.48 -> 35.87 ± 6.69 | Helped |
| average model weights rather than ensemble model outputs (ablation) | Multiplicative Weights (h=3) · CIFAR-10 | Min. Accuracy: 39.13 ± 1.74 -> 40.52 ± 1.28 | No clear effect |
| average model weights rather than ensemble model outputs (ablation) | Multiplicative Weights (h=3) · MNIST | Min. Acc.: 57.57 ± 8.46 -> 57.23 ± 8.51 | No clear effect |
Closest papers by meaning
- Adversarial robustness via robust low rank representations (NeurIPS 2020 · similarity 0.90)
- Adversarial Robustness of Supervised Sparse Coding (NeurIPS 2020 · similarity 0.90)
- Attribute-Guided Adversarial Training for Robustness to Natural Perturbations (AAAI 2021 · similarity 0.90)
- Adversarial Training and Robustness for Multiple Perturbations (NeurIPS 2019 · similarity 0.90)
- Adversarial Robustness against Multiple and Single $l_p$-Threat Models via Quick Fine-Tuning of Robust Classifiers (ICML 2022 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
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.