Fine-grained Video Action Understanding on FineBench Full Dataset v1.0
70.7Person MovementQwen2.5-VL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-VLSize=7B2026.05 | 70.7 | 63.8 | 73.9 | 68.8 | |
| InternVL-2.5Size=8B2026.05 | 66.8 | 62.1 | 78.1 | 67.1 | |
| InternVL-2.5Size=4B2026.05 | 61.4 | 58.6 | 78.1 | 63.3 | |
| Qwen2.5-VLSize=3B2026.05 | 58 | 57.5 | 73.2 | 60.5 | |
| mPlugOwl-2Size=7B2026.05 | 57.6 | 49.2 | 78.5 | 58.3 | |
| MiniCPM-2.6Size=8B2026.05 | 56.2 | 56.5 | 72.8 | 59.2 | |
| LLaVA-OVSize=7B2026.05 | 53.3 | 60.4 | 69.6 | 58.6 | |
| mPlugOwl-3Size=7B2026.05 | 48.9 | 54.8 | 75.2 | 55.6 | |
| SmolVLMSize=2B2026.05 | 47.9 | 50.5 | 71 | 52.9 | |
| BLIP-3Size=4B2026.05 | 34.3 | 58.6 | 64.9 | 48.2 | |
| InternVL-2.5Size=1B2026.05 | 33.8 | 40.2 | 79.6 | 44.1 |