Image Classification on MNIST (Absolute, Normalized, Average Accuracy)
84.1Average AccuracyFinetuned
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FinetunedBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Baseline2026.03 | 84.1 | 99.3 | — | |
| TARA-Variant BBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 76.3 | — | 82.7 | |
| TARA-Variant ABackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 74 | — | 76.3 | |
| Iso-CBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 73.6 | — | 68.9 | |
| Iso-CTSBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 73.5 | — | 68.5 | |
| AdaMergingBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 70 | — | 63 | |
| KnOTS-TIESBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 68 | — | 68.9 | |
| TIESBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 64.9 | — | 69 | |
| EMR-MergingBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 64 | — | 54.6 | |
| KnOTS-DARE-TIESBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 63.9 | — | 53.2 | |
| TABackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 63.8 | — | 53.4 | |
| LinearBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 63.7 | — | 55.3 | |
| RobustMergeBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 63.3 | — | 55.4 | |
| DARE-TIESBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 63.1 | — | 59.2 | |
| LoRA-LEGOBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 62.9 | — | 48.6 | |
| SVDBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=LoRA-aware Merging2026.03 | 62.6 | — | 52.8 | |
| RegMeanBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 60.9 | — | 43.1 | |
| FR-MergingBackbone=ViT-B/32, Pre-trained=ImageNet-21k, Fine-tuning Protocol=rank-16 LoRA, Merging Category=Vanilla Merging2026.03 | 57.6 | — | 46.6 |