Instance-level video understanding on V2P-Bench
94.3Overall AccuracyVideoSeeker-8B
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VideoSeeker-8BAgent=enabled2026.05 | 94.3 | 79.1 | 53.7 | 76.5 | 56.5 | 78.6 | 66.8 | 62.3 | 55 | 78.8 | 78.8 | 70.3 | 74.5 | |
| VideoSeeker-4BAgent=enabled2026.05 | 85.7 | 79.4 | 60.3 | 70.6 | 55.6 | 59.1 | 61.4 | 55.6 | 56 | 76.5 | 69.2 | 66.7 | 70.6 | |
| Gemini-2.5-ProAgent=disabled2026.05 | 84 | 72.4 | 68.2 | 71.8 | 75 | 73.3 | 22.6 | 66.7 | 72.7 | 47.4 | 67.5 | 63.6 | 69.8 | |
| Qwen3-VL-8BAgent=disabled2026.05 | 81.8 | 68.1 | 41.3 | 67.8 | 45.4 | 58.3 | 58.3 | 47.4 | 38.5 | 56.2 | 70.5 | 51.3 | 60.8 | |
| GPT-4oAgent=disabled2026.05 | 76.6 | 68.9 | 41.3 | 60.8 | 67 | 73.3 | 67.6 | 68.1 | 70.5 | 50 | 54 | 48.4 | 65.4 | |
| InternVL3-8BAgent=disabled2026.05 | 73.9 | 69.1 | 39.1 | 60.8 | 58.1 | 65.9 | 41 | 52.6 | 41.1 | 61.1 | 69.7 | 64.9 | 61.7 | |
| Qwen3-VL-4BAgent=disabled2026.05 | 73.7 | 67.1 | 41.2 | 58.8 | 51.7 | 52.9 | 51.2 | 42.1 | 42.2 | 63.2 | 71.9 | 48.6 | 59.2 | |
| Qwen3-VL-4BAgent=enabled2026.05 | 73 | 68.6 | 26.7 | 52.9 | 26.9 | 62.5 | 41 | 50 | 44.5 | 67.6 | 64.8 | 40.5 | 57.6 | |
| Qwen3-VL-8BAgent=enabled2026.05 | 71.1 | 71.2 | 39.1 | 58.8 | 48.3 | 41.2 | 64.3 | 42.1 | 51.7 | 57.9 | 68.4 | 59.4 | 59.9 | |
| LLaVA-OV-72BAgent=disabled2026.05 | 65.5 | 59.9 | 34.7 | 47 | 63.8 | 43.2 | 38.5 | 50 | 41.1 | 66.3 | 66.9 | 45.9 | 56.7 | |
| LLaVA-Video-72BAgent=disabled2026.05 | 62.2 | 60.8 | 30.4 | 54.9 | 61 | 54.5 | 43.6 | 47.4 | 42.2 | 70.5 | 71 | 59.5 | 58.6 | |
| LLaVA-Video-7BAgent=disabled2026.05 | 60.5 | 58.1 | 37 | 49 | 62.9 | 54.5 | 41 | 52.6 | 48.9 | 57.9 | 56.6 | 40.5 | 54.8 | |
| LLaVA-OV-7BAgent=disabled2026.05 | 57.1 | 52.1 | 28.3 | 47.1 | 63.8 | 59.1 | 41 | 42.1 | 35.6 | 63.2 | 62.8 | 43.2 | 52.8 |