Classification on SUN 397 (Top-1 Accuracy)
74Top-1 AccuracyPEcore
Evaluation Results
| Method | Links | |
|---|---|---|
| PEcoreData size=5B, Zero-shot=true2026.03 | 74 | |
| SigLIP2Data size=10B, Zero-shot=true2026.03 | 72.7 | |
| LlipData size=2.5B, Zero-shot=true2026.03 | 70.9 | |
| OpenCLIPData size=2B, Zero-shot=true2026.03 | 70 | |
| SigLIPData size=10B, Zero-shot=true2026.03 | 70 | |
| MetaCLIPData size=2.5B, Zero-shot=true2026.03 | 68.5 | |
| COSMOSData size=30M, Zero-shot=true2026.03 | 67.8 | |
| TPTBackbone=ViT-B/162026.06 | 67.21 | |
| GoldiCLIPData size=30M, Zero-shot=true2026.03 | 67.2 | |
| FLAIRData size=30M, Zero-shot=true2026.03 | 66.9 | |
| SigLIPData size=30M, Zero-shot=true2026.03 | 64 | |
| LSA+OnZetaBackbone=ViT-B/162026.06 | 63.93 | |
| DreamLIPData size=30M, Zero-shot=true2026.03 | 63.7 | |
| OnZetaBackbone=ViT-B/162026.06 | 63.65 | |
| LSABackbone=ViT-B/162026.06 | 63.12 | |
| CLIP baselineBackbone=ViT-B/162026.06 | 62.97 | |
| CLIPData size=30M, Zero-shot=true2026.03 | 62.2 | |
| TPTBackbone=ResNet-502026.06 | 61.46 | |
| LSA+OnZetaBackbone=ResNet-502026.06 | 53.62 | |
| OnZetaBackbone=ResNet-502026.06 | 53.27 | |
| LSABackbone=ResNet-502026.06 | 52.88 | |
| CLIP baselineBackbone=ResNet-502026.06 | 52.59 |