Research map / Label-efficient and robust learning / Recognition and long-tail
Why Do Unlearnable Examples Work: A Novel Perspective of Mutual Information
Yifan Zhu, Yibo Miao, Yinpeng Dong, Xiao-Shan Gao
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 |
|---|---|---|---|
| use MI-UE instead of prior UEs | unlearnable examples · CIFAR-10 | victim test accuracy (%): 11.25 -> 9.95 (vs TUE) | Improved |
| add the cosine similarity term to the MI-UE loss (ablation) | MI-UE · CIFAR-10 | victim test accuracy (%): 51.65 -> 9.95 | Improved |
| add the distance term to the MI-UE loss (ablation) | MI-UE · CIFAR-100 | victim test accuracy (%): 2.52 -> 1.17 | Improved |
| add the distance term to the MI-UE loss (ablation) | MI-UE · CIFAR-10 | victim test accuracy (%): 10.09 -> 9.95 | No clear effect |
| increase distance weight to 10 (ablation) | MI-UE · CIFAR-10 | victim test accuracy (%): 9.95 -> 45.90 | Hurt |
| add MINE-based MI regularizer to EM (ablation) | EM · CIFAR-10 | victim test accuracy (%): 24.17 -> 15.62 | Improved |
| add MINE-based MI regularizer to AP (ablation) | AP · CIFAR-10 | victim test accuracy (%): 11.21 -> 10.01 | Improved |
| increase poisoning epochs from 20 to 100 (ablation) | MI-UE · CIFAR-10 | victim test accuracy (%): 15.39 -> 9.95 | Improved |
| increase poison budget from 4/255 to 16/255 (ablation) | MI-UE · CIFAR-10 | victim test accuracy (%): 10.49 -> 9.97 | No clear effect |
| use MI-UE under adversarial training defense | unlearnable examples · CIFAR-10 (AT-6) | victim test accuracy (%): 86.82 -> 45.55 (vs SEM) | Improved |
| use MI-UE under AT-4 versus SEM | unlearnable examples · CIFAR-10 (AT-4) | victim test accuracy (%): 29.77 -> 31.79 | Hurt |
| use MI-UE against tailored defenses ISS and AVA | unlearnable examples · CIFAR-10 (ISS) | victim test accuracy (%): 80.84 (NTGA) vs 81.35 (MI-UE) | No clear effect |
Closest papers by meaning
- Detection and Defense of Unlearnable Examples (AAAI 2024 · similarity 0.92)
- How Far Are We from True Unlearnability? (ICLR 2025 · similarity 0.89)
- Ungeneralizable Examples (CVPR 2024 · similarity 0.88)
- Stable Unlearnable Example: Enhancing the Robustness of Unlearnable Examples via Stable Error-Minimizing Noise (AAAI 2024 · similarity 0.88)
- One for All: A Universal Generator for Concept Unlearnability via Multi-Modal Alignment (ICML 2024 · similarity 0.88)
- A3: Few-shot Prompt Learning of Unlearnable Examples with Cross-Modal Adversarial Feature Alignment (CVPR 2025 · similarity 0.87)
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.
