Research map / Architectures and efficiency / Quantization
Robust Machine Unlearning for Quantized Neural Networks via Adaptive Gradient Reweighting with Similar Labels
Yujia Tong, Yuze Wang, Jingling Yuan, Chuang Hu
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 Similar Labels (replace random labels for forgotten data) (ablation) | Q-MUL · CIFAR-100 10% (ResNet18) | Average Gap: 4.10 -> 3.11 | Improved |
| add Similar Labels (ablation) | Q-MUL · CIFAR-100 10% (ResNet18) | RA gap to Retrain: 3.52 -> 2.09 | Improved |
| add Adaptive Gradient Reweighting (ablation) | Q-MUL · CIFAR-100 10% (ResNet18) | Average Gap: 4.57 -> 3.11 | Improved |
| add Adaptive Gradient Reweighting (ablation) | Q-MUL · CIFAR-100 10% (ResNet18) | FA gap to Retrain: 4.92 -> 0.95 | Improved |
| use Q-MUL instead of SalUn | ResNet18 4-bit · CIFAR-100 10% random forgetting | Average Gap: 6.05 -> 3.11 | Improved |
| use Q-MUL instead of SalUn | ResNet18 4-bit · CIFAR-10 10% random forgetting | Average Gap: 1.31 -> 1.30 | No clear effect |
| use Q-MUL instead of SalUn | ResNet18 4-bit · CIFAR-100 50% random forgetting | Average Gap: 7.33 -> 7.06 | Improved |
| use Q-MUL instead of SalUn | ResNet18 PACT QAT · CIFAR-100 10% | Average Gap: 9.26 -> 6.00 | Improved |
| use Q-MUL instead of RL | ResNet18 DSQ QAT · CIFAR-100 10% | Average Gap: 9.88 -> 8.26 | Improved |
| use Q-MUL instead of l1-sparse | ResNet18 4-bit · Tiny-ImageNet 10% | Average Gap: 9.02 -> 8.17 | Improved |
| use Q-MUL instead of RL/l1-sparse/SalUn in class-wise forgetting | ResNet18 4-bit · CIFAR-10 class-wise | Average Gap: 0.13, 0.08, 0.10 vs 0.11 | No clear effect |
| use Q-MUL instead of SalUn | MobileNetV2 2-bit · CIFAR-100 10% | Average Gap: 9.29 -> 6.52 | Improved |
Closest papers by meaning
- LoTUS: Large-Scale Machine Unlearning with a Taste of Uncertainty (CVPR 2025 · similarity 0.89)
- When to Forget? Complexity Trade-offs in Machine Unlearning (ICML 2025 · similarity 0.88)
- Certified Unlearning for Neural Networks (ICML 2025 · similarity 0.88)
- Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models (ICLR 2024 · similarity 0.88)
- Certified Machine Unlearning via Noisy Stochastic Gradient Descent (NeurIPS 2024 · similarity 0.87)
- Unified Gradient-Based Machine Unlearning with Remain Geometry Enhancement (NeurIPS 2024 · 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.
