Continual Learning on UCIT (Last)
88.8ImageNet-R AccuracyOctopus
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| OctopusMethod Category=Regularization-based, Replay=No, Historical task proxy approximation=true2026.05 | 88.8 | 89.83 | 56.44 | 59.9 | 70.6 | 57.11 | 70.45 | |
| Vanilla RehearsalReplay=Yes2026.05 | 88.43 | 85.36 | 51.5 | 60.1 | 67.13 | 58.13 | 68.44 | |
| OctopusMethod Category=Regularization-based, Replay=No2026.05 | 88.2 | 93.03 | 58.46 | 60.5 | 69 | 56.87 | 71.01 | |
| HiDe-LLaVAMethod Category=Architecture-based, Replay=No2026.05 | 80.5 | 89.83 | 48.78 | 62.9 | 47.97 | 55.15 | 64.19 | |
| O-LoRAMethod Category=Regularization-based, Replay=No2026.05 | 69.36 | 82.42 | 48.64 | 53.66 | 42.53 | 53.52 | 58.36 | |
| MoELoRAMethod Category=Architecture-based, Replay=No2026.05 | 49.87 | 77.63 | 43.65 | 46.4 | 36.47 | 58.34 | 52.06 | |
| LwFMethod Category=Regularization-based, Replay=No2026.05 | 40.27 | 75.93 | 42.76 | 44.38 | 37.43 | 56.34 | 49.52 | |
| EWCMethod Category=Regularization-based, Replay=No2026.05 | 39.05 | 77.88 | 43.24 | 45.33 | 39.72 | 55.94 | 50.2 | |
| Sequential FinetuneReplay=No2026.05 | 37.63 | 72.33 | 43.47 | 41.7 | 35.63 | 57.95 | 48.12 | |
| L2PMethod Category=Architecture-based, Replay=No2026.05 | 32.73 | 80.41 | 43.72 | 42.16 | 39.25 | 52.77 | 48.51 |