Video Classification on EK100 (test val)
56.4Top-1 AccPrevious SoTA
Evaluation Results
| Method | Links | |
|---|---|---|
| Previous SoTAZero-shot=false2025.12 | 56.4 | |
| VL-JEPASFTArchitecture=ViT-L, Zero-shot=false2025.12 | 44.6 | |
| VL-JEPABASEArchitecture=ViT-L, Zero-shot=true2025.12 | 21.8 | |
| PE-CoreArchitecture=ViT-G, # Parameters=2.3B, Zero-shot=true2025.12 | 6.4 | |
| SigLIP2Architecture=ViT-g, # Parameters=1.9B, Zero-shot=true2025.12 | 6.1 | |
| PE-CoreArchitecture=ViT-L, # Parameters=671M, Zero-shot=true2025.12 | 6 | |
| SigLIP2Architecture=ViT-L, # Parameters=882M, Zero-shot=true2025.12 | 4.5 | |
| CLIPArchitecture=ViT-L, # Parameters=389M, Zero-shot=true2025.12 | 3.7 | |
| PE-CoreArchitecture=ViT-B, # Parameters=448M, Zero-shot=true2025.12 | 3.3 | |
| SigLIP2Architecture=ViT-B, # Parameters=375M, Zero-shot=true2025.12 | 2.3 | |
| CLIPArchitecture=RN50, # Parameters=75M, Zero-shot=true2025.12 | 1.5 | |
| CLIPArchitecture=ViT-B, # Parameters=124M, Zero-shot=true2025.12 | 1.3 |