Continual Learning on Opp1
27.27APVanilla LoRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Vanilla LoRArank=642026.05 | 27.27 | 12.53 | 14.74 | |
| LoRAMrank=642026.05 | 27.22 | 24.25 | 2.97 | |
| LoRA-GARank=1282026.05 | 26.56 | 26.62 | 0.06 | |
| Vanilla LoRARank=1282026.05 | 26.39 | 5.33 | 21.07 | |
| LoRAMRank=1282026.05 | 25.96 | 24.82 | 1.15 | |
| LoRA-GArank=642026.05 | 22.76 | 17.62 | 5.14 | |
| Slicerank=64, c=1.00, baseline=LoRA-GA2026.05 | 2.77 | 11.87 | -9.1 | |
| SliceRank=128, Base Initializer=LoRAM, c=1.002026.05 | 1.39 | 0.82 | 0.58 | |
| SliceRank=128, Base Initializer=LoRAM, c=0.502026.05 | 1.14 | 2.5 | 3.65 | |
| Slicerank=64, c=0.50, baseline=LoRA-GA2026.05 | 1.06 | 5.89 | -4.83 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=1.002026.05 | 0.96 | 20.31 | 19.34 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=1.002026.05 | 0.8 | 0.98 | 1.78 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=0.502026.05 | 0.71 | 16.99 | 16.28 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=0.502026.05 | 0.55 | 4.31 | 4.85 | |
| SliceRank=128, Base Initializer=LoRAM, c=0.752026.05 | 0.31 | 2.27 | 2.58 | |
| SliceRank=128, Base Initializer=LoRA-GA, c=0.752026.05 | 0.29 | 4.08 | 3.79 | |
| SliceRank=128, Base Initializer=Vanilla LoRA, c=0.752026.05 | 0.12 | 17.22 | 17.34 | |
| Slicerank=64, c=0.75, baseline=LoRA-GA2026.05 | -0.18 | 3.95 | -4.12 | |
| Slicerank=64, c=1.00, baseline=LoRAM2026.05 | -1.68 | 5.24 | -6.93 | |
| Slicerank=64, c=1.00, baseline=Vanilla LoRA2026.05 | -1.73 | 16.97 | -18.7 | |
| Slicerank=64, c=0.50, baseline=LoRAM2026.05 | -3.4 | -0.74 | -2.66 | |
| Slicerank=64, c=0.50, baseline=Vanilla LoRA2026.05 | -3.44 | 10.98 | -14.43 | |
| Slicerank=64, c=0.75, baseline=LoRAM2026.05 | -4.64 | -2.68 | -1.96 | |
| Slicerank=64, c=0.75, baseline=Vanilla LoRA2026.05 | -4.69 | 9.04 | -13.73 |