Continual Learning on Long Sequence Order 1 Qwen-3-4B
83.98APFPP + SABER-P
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FPP + SABER-PBackbone=Qwen-3-4B2026.05 | 83.98 | 2.14 | |
| SPA + SABER-PBackbone=Qwen-3-4B2026.05 | 81.97 | 1.87 | |
| SHLPTBackbone=Qwen-3-4B2026.05 | 80.95 | -0.27 | |
| ProgPromptBackbone=Qwen-3-4B2026.05 | 79.7 | -0.18 |