Continual Learning 벤치마크
Continual Learning on split CIFAR-100
Average Accuracy
- 2024-03-12 — OFSCIL: Average Accuracy 68.62
- 2024-11-21 — CODE-CL: Average Accuracy 77.21
| Rank | Model | Average Accuracy | BWT | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | CODE-CL | 77.21 | -1.1 | CODE-CL: Conceptor-Based Gradient Projection for Deep Continual Learning | mapolinario94/CODE-CL | 2024 |
| 2 | Continual 자동 추출 | 69.41 | – | CBPNet: A Continual Backpropagation Prompt Network for Alleviating Plasticity Loss on Edge Devices | 2025 | |
| 3 | OFSCIL | 68.62 | – | 12 mJ per Class On-Device Online Few-Shot Class-Incremental Learning | pulp-platform/fscil | 2024 |