Continual Learning on Opp2
27.73APLoRAM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LoRAMRank=1282026.05 | 27.73 | 30.05 | 2.31 | |
| Vanilla LoRArank=642026.05 | 26.5 | 25.1 | 1.4 | |
| LoRA-GArank=642026.05 | 24.87 | 16.51 | 8.37 | |
| LoRAMrank=642026.05 | 24.85 | 31.58 | -6.73 | |
| LoRA-GARank=1282026.05 | 24.81 | 26.24 | 1.43 | |
| Vanilla LoRARank=1282026.05 | 15.62 | 26.19 | 10.57 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=0.502026.05 | 10.23 | 0.34 | 9.89 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=1.002026.05 | 9.55 | 4.39 | 5.17 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=0.752026.05 | 9.48 | 4.02 | 5.46 | |
| SliceRank=128, Base Initializer=LoRAM, c=0.752026.05 | 2.63 | 0.16 | 2.79 | |
| SliceRank=128, Base Initializer=LoRAM, c=1.002026.05 | 2.55 | 0.53 | 3.09 | |
| Slicerank=64, c=0.75, baseline=LoRAM2026.05 | 2.11 | -14.82 | 16.92 | |
| Slicerank=64, c=0.75, baseline=LoRA-GA2026.05 | 2.08 | 0.26 | 1.82 | |
| SliceRank=128, Base Initializer=LoRAM, c=0.502026.05 | 1.88 | 3.51 | 1.64 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=0.502026.05 | 1.04 | 0.29 | 0.75 | |
| Slicerank=64, c=0.50, baseline=LoRAM2026.05 | 0.67 | -17.2 | 17.88 | |
| Slicerank=64, c=0.50, baseline=LoRA-GA2026.05 | 0.65 | -2.13 | 2.78 | |
| Slicerank=64, c=0.75, baseline=Vanilla LoRA2026.05 | 0.46 | -8.33 | 8.79 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=1.002026.05 | 0.37 | 4.34 | 3.97 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=0.752026.05 | 0.3 | 3.97 | 3.67 | |
| Slicerank=64, c=1.00, baseline=LoRAM2026.05 | -0.8 | -17.35 | 16.54 | |
| Slicerank=64, c=1.00, baseline=LoRA-GA2026.05 | -0.83 | -2.27 | 1.44 | |
| Slicerank=64, c=0.50, baseline=Vanilla LoRA2026.05 | -0.98 | -10.72 | 9.75 | |
| Slicerank=64, c=1.00, baseline=Vanilla LoRA2026.05 | -2.45 | -10.86 | 8.41 |