Video Understanding on MotionBench (dev)
58.3OverallQwen2.5-VL-72B*
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Qwen2.5-VL-72B*Frames=1fps2026.06 | 58.3 | 54.32 | 64 | 60.3 | 48.6 | 73.2 | 46.8 | 33 | |
| GLM4-9B + TE FusionFrames=162026.06 | 58 | — | — | — | — | — | — | — | |
| InternVL3-8B + MotionEnhancerFrames=82026.06 | 57.69 | 53.22 | 64.14 | 60.07 | 48.83 | 75.94 | 40.85 | 29.5 | |
| Qwen2.5-VL-7B + MotionEnhancerFrames=1fps2026.06 | 57.04 | 52.92 | 63.4 | 61.54 | 47.27 | 70.29 | 43.55 | 31.5 | |
| Qwen2.5-VL-3B + MotionEnhancerFrames=1fps2026.06 | 56.6 | 52.51 | 63.06 | 61.72 | 47.01 | 68.84 | 43.16 | 31.25 | |
| InternVL3-2B + MotionEnhancerFrames=82026.06 | 55.5 | 51.35 | 61.57 | 57.51 | 46.23 | 71.3 | 42 | 29.5 | |
| Qwen2.5-VL-7B + MotionSight*Frames=1fps2026.06 | 55.3 | 51.56 | 59.88 | 57.33 | 47.01 | 73.91 | 40.46 | 30.75 | |
| InternVL3-8B*Frames=82026.06 | 54.88 | 50.81 | 60.42 | 58.06 | 43.64 | 70.29 | 43.93 | 28.5 | |
| InternVL3-2B*Frames=82026.06 | 53.96 | 49.69 | 60.01 | 57.69 | 43.9 | 70 | 40.27 | 26.25 | |
| Qwen2.5-VL-3B*Frames=1fps2026.06 | 53.56 | 49.45 | 59.54 | 53.11 | 38.44 | 70.14 | 40.46 | 35 | |
| Qwen2.5-VL-7B*Frames=1fps2026.06 | 52.81 | 48.29 | 59 | 54.58 | 35.58 | 71.3 | 38.54 | 30.75 | |
| MiniCPM-V2.6-7BFrames=642026.06 | 52 | — | — | — | — | — | — | — | |
| PLLaVA-34BFrames=162026.06 | 52 | — | — | — | — | — | — | — |