Video Classification on SS v2 (test val)
77.5Top-1 AccuracyPrevious SoTA
Evaluation results
| Method | Links | |
|---|---|---|
| Previous SoTAZero-shot=false, Generalist Model=false2025.12 | 77.5 | |
| VL-JEPASFTArchitecture=ViT-L, # Parameters=1.6B, # Samples Seen=3.6B, Zero-shot=false, Generalist Model=true2025.12 | 73.2 | |
| VL-JEPABASEArchitecture=ViT-L, # Parameters=1.6B, # Samples Seen=3.3B, Zero-shot=true, Generalist Model=true2025.12 | 19.3 | |
| PE-CoreArchitecture=ViT-L, # Parameters=671M, # Samples Seen=58B, Zero-shot=true, Generalist Model=true2025.12 | 9.3 | |
| PE-CoreArchitecture=ViT-G, # Parameters=2.3B, # Samples Seen=86B, Zero-shot=true, Generalist Model=true2025.12 | 9 | |
| SigLIP2Architecture=ViT-g, # Parameters=1.9B, # Samples Seen=40B, Zero-shot=true, Generalist Model=true2025.12 | 6.1 | |
| SigLIP2Architecture=ViT-L, # Parameters=882M, # Samples Seen=40B, Zero-shot=true, Generalist Model=true2025.12 | 5.9 | |
| PE-CoreArchitecture=ViT-B, # Parameters=448M, # Samples Seen=58B, Zero-shot=true, Generalist Model=true2025.12 | 5.8 | |
| SigLIP2Architecture=ViT-B, # Parameters=375M, # Samples Seen=40B, Zero-shot=true, Generalist Model=true2025.12 | 5.2 | |
| CLIPArchitecture=ViT-L, # Parameters=389M, # Samples Seen=12.8B, Zero-shot=true, Generalist Model=true2025.12 | 3.8 | |
| CLIPArchitecture=ViT-B, # Parameters=124M, # Samples Seen=12.8B, Zero-shot=true, Generalist Model=true2025.12 | 3.1 | |
| CLIPArchitecture=RN50, # Parameters=75M, # Samples Seen=12.8B, Zero-shot=true, Generalist Model=true2025.12 | 2.1 |