Multi-Task Incremental Learning on MTIL benchmark
3Inference TimeDIKI
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DIKI2026.04 | 3 | 76.7 | |
| DPW (Ours†)2026.04 | 3 | 78.7 | |
| DPW (Ours)2026.04 | 3 | 79.3 | |
| MoE-Adapters2026.04 | 35 | 76.9 |
| Method | Links | ||
|---|---|---|---|
| DIKI2026.04 | 3 | 76.7 | |
| DPW (Ours†)2026.04 | 3 | 78.7 | |
| DPW (Ours)2026.04 | 3 | 79.3 | |
| MoE-Adapters2026.04 | 35 | 76.9 |