Continual Learning on Long Sequence Order 2 Qwen-3-4B
83.32Average Performance (AP)FPP + SABER-P
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FPP + SABER-PBackbone=Qwen-3-4B2026.05 | 83.32 | 1.96 | |
| SPA + SABER-PBackbone=Qwen-3-4B2026.05 | 82.68 | 2.12 | |
| SHLPTBackbone=Qwen-3-4B2026.05 | 80.54 | -0.23 | |
| ProgPromptBackbone=Qwen-3-4B2026.05 | 78.32 | -0.28 |