Research map / Continual learning / Online and replay-based
Fortuitous Forgetting in Connectionist Networks
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 |
|---|---|---|---|
| Add later-layer forgetting to CS-KD | ResNet18 · Flower | accuracy: 68.68 { $\pm 0.28$} -> 74.68 { $\pm 0.19$} | Helped |
| Add later-layer forgetting to CS-KD | ResNet18 · CUB | accuracy: 69.59 { $\pm 0.40$} -> 73.51 { $\pm 0.35$} | Helped |
| Add later-layer forgetting to CS-KD | ResNet18 · Aircraft | accuracy: 68.70 { $\pm 0.33$} -> 72.01 { $\pm 0.23$} | Helped |
| Add later-layer forgetting to CS-KD | ResNet18 · MIT67 | accuracy: 56.12 { $\pm 0.43$} -> 62.89 { $\pm 0.59$} | Helped |
| Add later-layer forgetting to CS-KD | ResNet18 · Stanford-Dogs | accuracy: 64.96 { $\pm 0.17$} -> 69.20 { $\pm 0.12$} | Helped |
| Add later-layer forgetting to label smoothing | ResNet50 · Tiny-ImageNet | accuracy: 49.27 -> 56.92 | Helped |
| Add later-layer forgetting to label smoothing | WideResNet-28-10 · CIFAR-10 | accuracy: 96.32 -> 95.91 | Hurt |
| Add later-layer forgetting to label smoothing | DenseNet-BC · CIFAR-100 | accuracy: 78.12 -> 78.13 | No clear effect |
| Reinitialize earlier layers instead of later layers (ablation) | LLF on Flower with ResNet18 · Flower | accuracy | Hurt |
| Freeze early layers after the initial generation (ablation) | LLF on Flower with ResNet18 · Flower | accuracy | No clear effect |
| Freeze reinitialized later layers during relearning (ablation) | LLF on Flower with ResNet18 · Flower | accuracy | Helped |
| Keep later layers frozen at the same initialization in every generation (ablation) | LLF on Flower with ResNet18 · Flower | accuracy | Hurt |
Closest papers by meaning
- Forgetting Has Neighbors: Localized Collateral Forgetting in Machine Unlearning (NeurIPS 2026 · similarity 0.89)
- Kolmogorov-Arnold Networks Still Catastrophically Forget but Differently from MLP (AAAI 2025 · similarity 0.89)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.88)
- Fast Machine Unlearning without Retraining through Selective Synaptic Dampening (AAAI 2024 · similarity 0.88)
- Measuring Representational Shifts in Continual Learning: A Linear Transformation Perspective (ICML 2025 · similarity 0.88)
- Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance (NeurIPS 2026 · similarity 0.88)
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.