Long Video Reasoning on LongVideoReason (eval)
80.3AccuracyVideoZoomer
Evaluation results
| Method | Links | |
|---|---|---|
| VideoZoomerSize=7B, Evaluation Protocol=Max 128 frames2025.12 | 80.3 | |
| LongVT-RL-7BCategory=Tool-calling Methods, Evaluated by Authors=true2026.05 | 75.1 | |
| VISDCategory=Ours2026.05 | 73.5 | |
| Video-R1Size=7B, Evaluation Protocol=Max 128 frames2025.12 | 72.8 | |
| Qwen2.5-VLSize=7B, Evaluation Protocol=Max 128 frames2025.12 | 70.8 | |
| VisionCoach-7BCategory=Tool-free Methods, Evaluated by Authors=true2026.05 | 70.7 | |
| VideoRFT-7BModel=VideoRFT-7B2025.10 | 69.4 | |
| Open-o3-Video-7B (Ours)Model=Open-o3-Video-7B (Ours)2025.10 | 69.4 | |
| VideoRFT-7BCategory=Tool-free Methods2026.05 | 69.4 | |
| Open-o3-Video-7BCategory=Tool-free Methods, Evaluated by Authors=true2026.05 | 69.1 | |
| VideoR1-7BModel=VideoR1-7B2025.10 | 68.9 | |
| VideoR1-7BCategory=Tool-free Methods2026.05 | 68.9 | |
| LongVILA-R1Size=7B, Evaluation Protocol=Standard2025.12 | 67.9 | |
| Gemini-1.5-ProSize=-, Evaluation Protocol=Standard2025.12 | 67.3 | |
| GPT-4oCategory=Proprietary Models2026.05 | 66 | |
| InternVL-2.5-8BModel=InternVL-2.5-8B2025.10 | 62 | |
| GPT-4oSize=-, Evaluation Protocol=Standard2025.12 | 60.7 | |
| VideoLLaMA3-7BModel=VideoLLaMA3-7B2025.10 | 59.8 | |
| Qwen2.5-VL-7B (Base)Model=Qwen2.5-VL-7B (Base)2025.10 | 59.3 | |
| Qwen2.5-VL-7BCategory=Tool-free Methods2026.05 | 59.3 |