Research map / Continual learning / Continual fine-tuning of LLMs
Boundary Unlearning: Rapid Forgetting of Deep Networks via Shifting the Decision Boundary
Read the paperCheck 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 |
|---|---|---|---|
| Apply Boundary Shrink to forget one class | All-CNN · CIFAR-10 Df | accuracy: 99.92 -> 5.94 | Helped |
| Apply Boundary Shrink to forget one class | ResNet50 · Vggface2 Df | accuracy: 98.57 -> 1.54 | Helped |
| Apply Boundary Shrink while retaining remaining-class utility | All-CNN · CIFAR-10 Dr | accuracy: 99.97 -> 99.24 | No clear effect |
| Apply Boundary Expanding to forget one class | All-CNN · CIFAR-10 Df | accuracy: 99.92 -> 8.96 | Helped |
| Apply Boundary Expanding to forget one class | ResNet50 · Vggface2 Df | accuracy: 98.57 -> 4.22 | Helped |
| Use Boundary Shrink rather than Boundary Expanding | Boundary Unlearning · CIFAR-10 Df | accuracy: 5.94 vs 8.96 | Mixed |
| Use Boundary Shrink rather than Boundary Expanding | Boundary Unlearning · Vggface2 Df | accuracy: 1.54 vs 4.22 | Mixed |
| Use Boundary Expanding rather than Boundary Shrink | Boundary Unlearning · CIFAR-10 unlearning runtime | speed-up over retraining: 29.7× vs 4.8× | Helped |
| Use Boundary Expanding rather than Boundary Shrink | Boundary Unlearning · Vggface2 unlearning runtime | speed-up over retraining: 26.6× vs 11.7× | Helped |
| Use Fisher Forgetting instead of Boundary Unlearning | CIFAR-10 model · CIFAR-10 Dr | accuracy: 61.62 | Hurt |
Closest papers by meaning
- Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models (ICLR 2024 · similarity 0.89)
- When to Forget? Complexity Trade-offs in Machine Unlearning (ICML 2025 · similarity 0.88)
- Distribution-Level Feature Distancing for Machine Unlearning: Towards a Better Trade-off Between Model Utility and Forgetting (AAAI 2025 · similarity 0.88)
- Not All Wrong is Bad: Using Adversarial Examples for Unlearning (ICML 2025 · similarity 0.88)
- Partially Blinded Unlearning: Class Unlearning for Deep Networks from Bayesian Perspective (AAAI 2025 · similarity 0.88)
- 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.