Vision Model Merging on Vision Model Merging 8 tasks
95.8Average Absolute AccuracyFine-tuned
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Fine-tunedBackbone=ViT-L/142026.06 | 95.8 | — | |
| ESM++Backbone=ViT-L/14, Merging Category=Dynamic Merging, Rank (r)=322026.06 | 95.6 | 99.8 | |
| PACT-Iso-CBackbone=ViT-L/142026.06 | 95.1 | 98.3 | |
| Iso-CTSBackbone=ViT-L/142026.06 | 94.7 | 98.8 | |
| ESMBackbone=ViT-L/14, Merging Category=Static Merging2026.06 | 94.7 | 98.8 | |
| Fine-tunedBackbone=ViT-B/162026.06 | 94.6 | — | |
| PACT-TSV-MBackbone=ViT-L/142026.06 | 94.4 | 98.5 | |
| Iso-CBackbone=ViT-L/142026.06 | 94.2 | 98.3 | |
| ESM++Backbone=ViT-B/16, Merging Category=Dynamic Merging, Rank (r)=322026.06 | 93.3 | 98.6 | |
| TSV-MBackbone=ViT-L/142026.06 | 93 | 97 | |
| Fine-tunedBackbone=ViT-B/322026.06 | 92.8 | — | |
| PACT-TABackbone=ViT-L/142026.06 | 92.8 | 96.9 | |
| Fine-tunedBackbone=ViT-B/322026.06 | 92.8 | — | |
| IndividualBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 92.36 | — | |
| PACT-Iso-CBackbone=ViT-B/162026.06 | 92 | 97.2 | |
| WEMoEBackbone=ViT-B/32, Merging Category=Dynamic Merging2026.06 | 91.9 | 99 | |
| ESM++Backbone=ViT-B/32, Merging Category=Dynamic Merging, Rank (r)=322026.06 | 91.8 | 99 | |
| E-WEMoE-90%Backbone=ViT-B/32, Merging Category=Dynamic Merging2026.06 | 91.7 | 98.8 | |
| ESMBackbone=ViT-B/16, Merging Category=Static Merging2026.06 | 91.6 | 96.7 | |
| SMILEBackbone=ViT-B/32, Merging Category=Dynamic Merging2026.06 | 91.5 | 98.4 | |
| ESM++Backbone=ViT-B/32, Merging Category=Dynamic Merging, Rank (r)=82026.06 | 91.3 | 98.4 | |
| PACT-TSV-MBackbone=ViT-B/162026.06 | 91.2 | 96.3 | |
| Iso-CTSBackbone=ViT-B/162026.06 | 91.1 | 96.1 | |
| Iso-CBackbone=ViT-B/162026.06 | 90.6 | 95.6 | |
| IndividualBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 89.71 | — | |
| TSV-MBackbone=ViT-B/162026.06 | 89 | 93.9 | |
| ESMBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 88.6 | 95.4 | |
| PACT-TSV-MBackbone=ViT-B/322026.06 | 88.1 | 94.8 | |
| IndividualBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 87.82 | — | |
| PACT-Iso-CBackbone=ViT-B/322026.06 | 87.3 | 94.1 | |
| DC-MergeBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 87.1 | 93.6 | |
| TIESBackbone=ViT-L/142026.06 | 86.9 | 90.7 | |
| Consensus TABackbone=ViT-L/142026.06 | 86.3 | 90.1 | |
| Iso-CBackbone=ViT-B/322026.06 | 86.3 | 92.9 | |
| Iso-CBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 86.3 | 92.9 | |
| Iso-CTSBackbone=ViT-B/322026.06 | 86.2 | 92.8 | |
| Iso-CTSBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 86.2 | 92.8 | |
| TSV-MBackbone=ViT-B/322026.06 | 85.9 | 92.3 | |
| TSV-MBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 85.9 | 92.3 | |
| FREE-MergingBackbone=ViT-B/32, Merging Category=Dynamic Merging2026.06 | 85.8 | 92.4 | |
| PACT-TABackbone=ViT-B/162026.06 | 85.5 | 90.3 | |
| Task ArithmeticBackbone=ViT-L/142026.06 | 84.9 | 88.7 | |
| DC-MergeBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 82.61 | 89.42 | |
| Iso-CTSBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 81.64 | 88.31 | |
| TIESBackbone=ViT-B/162026.06 | 79.7 | 84.3 | |
| Weight AveragingBackbone=ViT-L/142026.06 | 79.6 | 83.2 | |
| Consensus TABackbone=ViT-B/162026.06 | 79.4 | 83.9 | |
| FR-MergingBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 78.6 | 84.6 | |
| TSV-MBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 76.52 | 83 | |
| Task ArithmeticBackbone=ViT-B/162026.06 | 75.4 | 79.6 | |
| PACT-TABackbone=ViT-B/322026.06 | 75.2 | 81 | |
| TIESBackbone=ViT-B/322026.06 | 75.1 | 81 | |
| TIES-MergingBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 75.1 | 81 | |
| Consensus TABackbone=ViT-B/322026.06 | 75 | 80.8 | |
| Consensus TABackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 75 | 80.8 | |
| KnOTS-TIESBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 73.61 | 79.92 | |
| Weight AveragingBackbone=ViT-B/162026.06 | 72.2 | 76.6 | |
| Task ArithmeticBackbone=ViT-B/322026.06 | 70.8 | 76.5 | |
| Task ArithmeticBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 70.8 | 76.5 | |
| DC-MergeBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 70.53 | 78.86 | |
| WUDI-MergingBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 69.78 | 75.91 | |
| Iso-CTSBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 69.06 | 77.38 | |
| Task ArithmeticBackbone=ViT-L-14, Fine-tuning setting=LoRA2026.03 | 68.29 | 74.38 | |
| Weight AveragingBackbone=ViT-B/322026.06 | 66.3 | 72.1 | |
| Model SoupBackbone=ViT-B/32, Merging Category=Static Merging2026.06 | 66.3 | 72.1 | |
| Zero-shotBackbone=ViT-L/142026.06 | 64.7 | — | |
| DC-MergeBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 64.17 | 73.9 | |
| Iso-CTSBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 63.01 | 72.71 | |
| TSV-MBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 62.97 | 70.87 | |
| KnOTS-TIESBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 60.8 | 68.59 | |
| WUDI-MergingBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 58.95 | 66.63 | |
| TSV-MBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 58.91 | 68.25 | |
| Task ArithmeticBackbone=ViT-B-16, Fine-tuning setting=LoRA2026.03 | 57.7 | 65.3 | |
| KnOTS-TIESBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 55.93 | 65.03 | |
| Zero-shotBackbone=ViT-B/162026.06 | 55.3 | — | |
| WUDI-MergingBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 55.25 | 64.38 | |
| Task ArithmeticBackbone=ViT-B-32, Fine-tuning setting=LoRA2026.03 | 52.8 | 61.73 | |
| Zero-shotBackbone=ViT-B/322026.06 | 48.3 | — | |
| Pre-trainedBackbone=ViT-B/322026.06 | 48.3 | — |