Continual Learning on Split-MNIST (train test)
98.9AccuracyHyperNet Learned
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HyperNet LearnedParadigm=Cond. Regen. (Cctx ≫ 0)2026.03 | 98.9 | 0 | 95.2 | — | |
| HyperNet OracleParadigm=Cond. Regen. (Cctx ≫ 0)2026.03 | 98.8 | 0 | 97.6 | — | |
| CFlow (ODE)Paradigm=State Transform. (Cctx ≈ 0)2026.03 | 92.4 | 6.1 | 0 | — | |
| Experience ReplayParadigm=Replay (Cctx = 0)2026.03 | 85.9 | 12.5 | 0 | — | |
| LwFParadigm=State Protection (Cctx = 0)2026.03 | 24.2 | 54.8 | 0 | — | |
| EWCParadigm=State Protection (Cctx = 0)2026.03 | 18.9 | 97.6 | 0 | — | |
| NaiveSGDParadigm=State Protection (Cctx = 0)2026.03 | 18.7 | 97.1 | 0 | — | |
| SIParadigm=State Protection (Cctx = 0)2026.03 | 16.4 | 97.1 | 0 | — |