Continual Reinforcement Learning on MiniGrid
70.6Mean SRTELAPA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TELAPA2026.04 | 70.6 | 3.35 | 30.3 | 50 | 62.9 | 25 | |
| FINETUNE2026.04 | 58.8 | 6.55 | 26.2 | 37 | 69.5 | 14 | |
| FINETUNE-RESET2026.04 | 57.6 | 5.91 | 26.3 | 33 | 70.9 | 15 | |
| SCR.-TRANSFER2026.04 | 52.5 | 5.49 | 5.4 | 30 | 0.9 | 24 | |
| TELAPA-STATICAblation=fixed embedder, disabled online re-embedding2026.04 | 50.7 | 5.59 | 29.4 | 38 | 70.2 | 12 | |
| S&P2026.04 | 46.3 | 6.86 | 33 | 33 | 84.4 | 6 | |
| DFF2026.04 | 38.8 | 8.04 | 10.1 | 9 | 21.2 | 1 | |
| SCRATCH2026.04 | 37.4 | 7.75 | 28.3 | 33 | 74.8 | 2 | |
| EWC2026.04 | 14.3 | 9.67 | 9 | 5 | 11.5 | 5 | |
| L2INIT2026.04 | 12 | 9.74 | 7.2 | 8 | 14 | 2 |