Fine-grained temporal action understanding with objects
Accuracy is the reported evaluation metric for Something-Something V2. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.
Higher is better
Muted rows were not state of the art when published — an earlier or same-year result already scored better.
| Rank | Model | Trust | Score | Year | Links | Fix |
|---|---|---|---|---|---|---|
| 01 | V-JEPA 2 ViT-g (1B, 384px) | unverified | 77.3 | 2025 | Paper ↗Code ↗ | Looks wrong? |
| 02 | VideoMAE ViT-L | unverified | 75.4 | 2022 | Paper ↗Code ↗ | Looks wrong? |
| 03 | DINOv3 (7B) | unverified | 70.8 | 2025 | Paper ↗Code ↗ | Looks wrong? |
| 04 | VideoPrism-g | unverified | 68.5 | 2024 | Paper ↗Code ↗ | Looks wrong? |
| 05 | DINOv2 (ViT-g/14) | unverified | 38.3 | 2023 | Paper ↗Code ↗ | Looks wrong? |