Video Action Recognition on MMEB Video zero-shot
63Overall AccuracyX-Ray visual
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| X-Ray visualBackbone=400M, Zero-shot=true2026.02 | 63 | 85.5 | 41.8 | 66.7 | 57.8 | 54.2 | |
| Qwen2 VLBackbone=7 Billion, Zero-shot=true2026.02 | 57.6 | 78.6 | 37.2 | 55.6 | 63.9 | 53 |