Research map / Label-efficient and robust learning / Recognition and long-tail
Stake the Points: Structure-Faithful Instance Unlearning
Kiseong Hong, JungKyoo Shin, Eunwoo 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 |
|---|---|---|---|
| add structure-aware alignment (ablation) | STRUCTGUARD · CIFAR-10 (k=256) | Atest accuracy: 37.75 -> 56.32 | Improved |
| add structure-aware alignment (ablation) | STRUCTGUARD · CIFAR-100 (k=256) | Atest accuracy: 51.49 -> 56.91 | Improved |
| add structure-aware regularization (ablation) | STRUCTGUARD · CIFAR-10 (k=256) | Atest accuracy: 55.01 -> 56.32 | Improved |
| add classifier regularization (ablation) | STRUCTGUARD · CIFAR-10 (k=256) | Atest accuracy: 54.08 -> 56.32 | Improved |
| add structure-aware regularization (ablation) | STRUCTGUARD · CIFAR-100 (k=256) | Atest accuracy: 55.86 -> 56.91 | Improved |
| add classifier regularization (ablation) | STRUCTGUARD · CIFAR-100 (k=256) | Atest accuracy: 56.42 -> 56.91 | Improved |
| use LLM-attribute semantic anchors instead of visual prototype anchors (ablation) | STRUCTGUARD · CIFAR-10 (k=256) | Atest accuracy: 48.78 -> 56.32 | Improved |
| use LLM-attribute semantic anchors instead of visual prototype anchors (ablation) | STRUCTGUARD · ImageNet-1K (k=256) | Atest accuracy: 40.95 -> 41.15 | Improved |
| use cosine similarity instead of MSE for alignment (ablation) | STRUCTGUARD · CIFAR-10 (k=256) | Atest accuracy: 43.80 -> 56.32 | Improved |
| use cosine similarity instead of Wasserstein distance for alignment (ablation) | STRUCTGUARD · CIFAR-100 (k=256) | Atest accuracy: 55.51 -> 56.91 | Improved |
| StructGuard (structure-preserving anchors) vs L2UL | StructGuard vs L2UL · ImageNet-1K (k=256) | Atest accuracy: 31.19 -> 41.15 | Improved |
| StructGuard vs ORACLE at large forget set | StructGuard vs ORACLE · CIFAR-100 (k=256) | Atest accuracy: 65.31 -> 56.91 | Hurt |
Closest papers by meaning
- The Unseen Threat: Residual Knowledge in Machine Unlearning under Perturbed Samples (NeurIPS 2025 · similarity 0.88)
- From Dormant to Deleted: Tamper-Resistant Unlearning Through Weight-Space Regularization (NeurIPS 2025 · similarity 0.86)
- Benchmarking Vision Language Model Unlearning via Fictitious Facial Identity Dataset (ICLR 2025 · similarity 0.86)
- Certified Unlearning for Neural Networks (ICML 2025 · similarity 0.86)
- SHRED: Retain-Set-Free Unlearning via Self-Distillation with Logit Demotion (NeurIPS 2026 · similarity 0.86)
- Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models (ICLR 2024 · similarity 0.86)
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.
