Image Classification on CIFAR-100 (full)
83.05AccuracyGLMC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GLMCBackbone=ResNet-502023.05 | 83.05 | — | |
| PaCoBackbone=ResNet-502023.05 | 79.1 | — | |
| CutMixBackbone=ResNet-502023.05 | 78.03 | — | |
| MixUpBackbone=ResNet-502023.05 | 77.99 | — | |
| SupConBackbone=ResNet-502023.05 | 76.5 | — | |
| vanillaBackbone=ResNet-502023.05 | 75.28 | — | |
| Hierarchy-CLIPZero-shot=true, Backbone=ViT-B/162022.12 | 35.2 | — | |
| CLIPZero-shot=true, Backbone=ViT-B/162022.12 | 31.8 | — | |
| FPET-LoRABackbone=ViT-B/162025.03 | 0.9229 | — | |
| FPET-AdaptFormerBackbone=ViT-B/162025.03 | 0.9218 | — | |
| LoRABackbone=ViT-B/162025.03 | 0.9217 | — | |
| AdaptFormerBackbone=ViT-B/162025.03 | 0.9214 | — | |
| Full-FTModel Architecture=ViT-B, # Parameters=86M2026.04 | — | 87.7 | |
| Full-FTModel Architecture=ViT-L, # Parameters=303M2026.04 | — | 87.7 | |
| LoRA (r=2)Model Architecture=ViT-L, # Parameters=98K2026.04 | — | 92.9 | |
| LoRA (r=4)Model Architecture=ViT-B, # Parameters=74K2026.04 | — | 90.3 | |
| LoRA (r=4)Model Architecture=ViT-L, # Parameters=197K2026.04 | — | 92.4 | |
| NOLAModel Architecture=ViT-B, # Parameters=48K2026.04 | — | 86.6 | |
| NOLAModel Architecture=ViT-L, # Parameters=96K2026.04 | — | 89.3 | |
| SOLAR NOLA(4→1.2)Model Architecture=ViT-B, # Parameters=32K2026.04 | — | 85.4 | |
| SOLAR NOLA(4→1.2)Model Architecture=ViT-L, # Parameters=64K2026.04 | — | 89.3 | |
| SOLAR r=2(1→0.3)Model Architecture=ViT-L, # Parameters=50K2026.04 | — | 90 | |
| SOLAR r=4(4→1.6)Model Architecture=ViT-B, # Parameters=41K2026.04 | — | 89.8 | |
| SOLAR r=4(4→1.6)Model Architecture=ViT-L, # Parameters=82K2026.04 | — | 91.4 |