Continual Learning on MNIST Split
99.42Avg AccuracyHEM-
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HEM-Strategy=LwF2025.01 | 99.42 | — | |
| MMStrategy=SI2025.01 | 98.76 | — | |
| CEStrategy=SI2025.01 | 98.68 | — | |
| HEM-Strategy=Reply2025.01 | 98.59 | — | |
| HEM-Strategy=EWC2025.01 | 98.1 | — | |
| CEStrategy=LwF2025.01 | 98.05 | — | |
| CEStrategy=Reply2025.01 | 97.56 | — | |
| MMStrategy=LwF2025.01 | 97.04 | — | |
| MMStrategy=Reply2025.01 | 96.44 | — | |
| HEM-Strategy=SI2025.01 | 96.36 | — | |
| CEStrategy=EWC2025.01 | 93.27 | — | |
| MoBseeds=5, allocation=VCG auction2025.11 | 88.77 | 11.9 | |
| LNStrategy=SI2025.01 | 87.74 | — | |
| MMStrategy=EWC2025.01 | 80.31 | — | |
| LNStrategy=Reply2025.01 | 63.86 | — | |
| LNStrategy=LwF2025.01 | 61.7 | — | |
| LNStrategy=EWC2025.01 | 53.65 | — | |
| DICEStrategy=LwF2025.01 | 50.89 | — | |
| DICEStrategy=EWC2025.01 | 50.53 | — | |
| DICEStrategy=Reply2025.01 | 50.17 | — | |
| DICEStrategy=SI2025.01 | 50.17 | — | |
| Random Assignmentseeds=52025.11 | 46.24 | 11.93 | |
| Monolithic EWCseeds=5, regularization=EWC2025.11 | 27.96 | 85.78 | |
| Naive Fine-tuningseeds=52025.11 | 19.8 | 99.81 | |
| Gated MoEseeds=52025.11 | 19.54 | 99.3 |