Research map / Adversarial robustness and security / Certified and empirical robustness
FUSE: Full‑spectrum Unlearnable Examples via Spectral Equalization
Jiale Cai, Gezheng Xu, Zhihao Li, Ruiyi Fang, Ruizhi Pu, Di Wu, Qicheng Lao, Charles Ling, Boyu Wang
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 |
|---|---|---|---|
| add Random Spectral Masking (ablation) | FUSE · CIFAR-10 low-pass 0.5, ResNet-18 | test accuracy: 57.35 -> 10.13 | Improved |
| add entropy regularizer (ablation) | FUSE · CIFAR-10 low-pass 0.5, ResNet-18 | test accuracy: 15.13 -> 10.13 | Improved |
| add Cross-Band Guidance (ablation) | FUSE · CIFAR-10 low-pass 0.5, ResNet-18 | test accuracy: 17.00 -> 10.13 | Improved |
| add Random Spectral Masking, unfiltered-like cutoff 1.0 (ablation) | FUSE · CIFAR-10 cutoff 1.0, ResNet-18 | test accuracy: 11.13 -> 9.41 | Improved |
| set frequency split radius r_c to 0.5 instead of 0.2 (ablation) | FUSE · CIFAR-10 low-pass 0.5 | test accuracy: 10.93 -> 10.13 | Improved |
| set frequency split radius r_c to 0.5 instead of 0.8 (ablation) | FUSE · CIFAR-10 low-pass 0.5 | test accuracy: 14.70 -> 10.13 | Improved |
| use moderate lambda around 0.5 rather than very small or very large (ablation) | FUSE · CIFAR-10, CIFAR-100, SVHN | test accuracy | Improved |
| increase perturbation bound from 2/255 to 8/255 (ablation) | FUSE · CIFAR-10 unfiltered | test accuracy: 11.00 -> 9.41 | Improved |
| increase perturbation bound from 2/255 to 8/255 under low-pass (ablation) | FUSE · CIFAR-10 low-pass 0.5 | test accuracy: 15.53 -> 10.13 | Improved |
| apply low-pass filtering to protected data | LSP · CIFAR-10 ResNet-18 | test accuracy: 13.54 -> 54.16 | Hurt |
| apply low-pass filtering to protected data | FUSE · CIFAR-10 ResNet-18 | test accuracy: 9.41 -> 10.13 | No clear effect |
| apply Mixup augmentation defense | OPS · CIFAR-10 ResNet-18 | test accuracy: 15.56 -> 33.13 | Hurt |
Later papers that compared with this method
1 later accepted paper reports Fuse as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Robust onion: Peeling Open Vocab Object Detectors Under Noise (ECCV 2026 · better on 4 of 5 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Detecting and Corrupting Convolution-based Unlearnable Examples (AAAI 2025 · similarity 0.86)
- How Far Are We from True Unlearnability? (ICLR 2025 · similarity 0.86)
- UnSeg: One Universal Unlearnable Example Generator is Enough against All Image Segmentation (NeurIPS 2024 · similarity 0.85)
- Modulate Your Spectrum in Self-Supervised Learning (ICLR 2024 · similarity 0.84)
- Purify Unlearnable Examples via Rate-Constrained Variational Autoencoders (ICML 2024 · similarity 0.84)
- One for All: A Universal Generator for Concept Unlearnability via Multi-Modal Alignment (ICML 2024 · 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.
