Research map / Adversarial robustness and security / Certified and empirical robustness
Angular Gradient Sign Method: Uncovering Vulnerabilities in Hyperbolic Networks
Minsoo Jo, Dongyoon Yang, Taesup Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| keep only angular component of feature shift (AGSM) instead of full FGSM | FGSM · CIFAR-100 PRN-32 | top-1 accuracy: 19.67 -> 13.93 | Improved |
| apply only radial shift of representation (ablation) | FGSM perturbation · CIFAR-100 PRN-32 | top-1 accuracy: 53.44 -> 53.44 | No clear effect |
| apply only angular shift of representation (ablation) | FGSM perturbation · CIFAR-100 PRN-32 | top-1 accuracy: 53.44 -> 25.56 | Improved |
| angular gradient (AGSM) vs FGSM | FGSM · CIFAR-10 PRN-32 eps3 | robust accuracy: 54.19 -> 41.56 | Improved |
| angular multi-step (PAGD) vs PGD | PGD · CIFAR-10 PRN-20 eps2 | robust accuracy: 21.40 -> 14.25 | Improved |
| PAGD vs PGD at large epsilon | PGD · CIFAR-100 PRN-20 eps3 | robust accuracy: 9.43 -> 9.43 | No clear effect |
| angular gradient (AGSM) vs FGSM | FGSM · Tiny ImageNet PRN-32 eps3 | robust accuracy: 8.02 -> 5.57 | Improved |
| angular gradient (AGSM) vs FGSM | FGSM · COCO T2I R@5 HyCoCLIP ViT-B/16 8.0/255 | recall@5: 10.80 -> 7.60 | Improved |
| angular multi-step (PAGD) vs PGD | PGD · COCO I2T R@5 HyCoCLIP ViT-B/16 8.0/255 | recall@5: 2.40 -> 1.90 | Improved |
| angular gradient (AGSM) vs FGSM | FGSM · Flickr30K I2T R@5 ViT-B/16 3.2/255 | recall@5: 29.50 -> 26.60 | Improved |
| angular gradient (AGSM) vs FGSM | FGSM · COCO hyperbolic distance 3.2/255 | mean distance: 0.3058 -> 0.3639 | Improved |
| angular gradient (AGSM) vs FGSM | FGSM · CIFAR-10 MSP drop 8.0/255 | confidence drop: 0.4364 -> 0.5597 | Improved |
Closest papers by meaning
- HHA: Hierarchical Hyperbolic Constraints for Imperceptible Point Cloud Attacks (ECCV 2026 · similarity 0.88)
- Geometry-Aware Generation of Adversarial Point Clouds (TPAMI 2020 · similarity 0.87)
- HyperDefender: A Robust Framework for Hyperbolic GNNs (AAAI 2025 · similarity 0.86)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.85)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.85)
- AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples (AAAI 2025 · 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.
