Image Classification on DOTA IN21K transfer
93.9Head AccuracyLoRA + MetaPEFT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LoRA + MetaPEFTCategory=Additive, Optimization=MetaPEFT2026.03 | 93.9 | 95.1 | 91.4 | |
| AdapterCategory=Additive2026.03 | 93.2 | 94.9 | 90.6 | |
| Adapter + MetaPEFTCategory=Additive, Optimization=MetaPEFT2026.03 | 93.2 | 95.1 | 90.7 | |
| AdaptFormerCategory=Additive2026.03 | 93.2 | 94.9 | 90.1 | |
| LoRACategory=Additive2026.03 | 93.1 | 93.4 | 90.7 | |
| AdaptFormer + MetaPEFTCategory=Additive, Optimization=MetaPEFT2026.03 | 92.7 | 94.6 | 90.1 | |
| VPT-DeepCategory=Non-additive2026.03 | 92.1 | 93.5 | 90.2 | |
| BitFitCategory=Non-additive2026.03 | 92 | 93.7 | 89.1 | |
| VPT-ShallowCategory=Non-additive2026.03 | 85.4 | 89 | 82.4 |