Video Backward Reasoning on OVO-Bench Backward
61.95EPMQwen3-VL-8B + OASIS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-VL-8B + OASISframes=-2026.04 | 61.95 | 60.14 | 47.31 | 57.21 | |
| Gemini 1.5 Proframes=1fps2026.04 | 58.59 | 76.35 | 52.64 | 62.54 | |
| GPT-4oframes=642026.04 | 57.91 | 75.68 | 48.66 | 60.75 | |
| LLaVA-OneVision-7Bframes=642026.04 | 56.24 | 55.41 | 21.51 | 43.71 | |
| LLaVA-Video-7Bframes=642026.04 | 56.23 | 53.47 | 51.53 | 53.74 | |
| Qwen2-VL-72Bframes=642026.04 | 52.53 | 60.81 | 57.53 | 56.95 | |
| Qwen2.5-VL-7B + OASISframes=-2026.04 | 51.85 | 58.78 | 48.92 | 52.61 | |
| Qwen3-VL-8Bframes=0.5fps2026.04 | 51.18 | 60.81 | 43.55 | 51.19 | |
| StreamChatframes=15fps2026.04 | 50.17 | 54.05 | 37.63 | 47.39 | |
| Qwen2.5-VL-7Bframes=0.5fps2026.04 | 49.83 | 58.11 | 44.62 | 50.24 | |
| Qwen2-VL-7Bframes=642026.04 | 48.71 | 35.45 | 45.03 | 43.73 | |
| Dispiderframes=1fps2026.04 | 48.48 | 55.41 | 45.43 | 36.06 | |
| Flash-VStream-7Bframes=1fps2026.04 | 39.06 | 37.16 | 5.91 | 27.38 | |
| VideoLLM-online-8Bframes=2fps2026.04 | 18.8 | 12.18 | 17.73 | 16.24 |