Video Question Answering on NA-VQA (Far)
67.64ComprehensionClaude Sonnet 4.5
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Claude Sonnet 4.5Type=Proprietary, Use Descr.=true2026.03 | 67.64 | 70.61 | 66.95 | 64.24 | |
| Qwen-235BType=Proprietary, Use Descr.=true2026.03 | 54.4 | 61.51 | 53.98 | 54.32 | |
| GPT-120B OSSType=Proprietary, Use Descr.=true2026.03 | 51.89 | 57.18 | 50.31 | 50.04 | |
| Video-NaRAType=Fine-Tuned Model, Use Descr.=true, Frames=1282026.03 | 35.98 | 39.88 | 35.1 | 33.67 | |
| Video-NaRAType=Zero-Shot VLM, Use Descr.=true, Frames=1282026.03 | 35.1 | 39.38 | 34.58 | 33.35 | |
| Video-LLaMA3 (7B)Type=Zero-Shot VLM, Frames=1282026.03 | 34.94 | 38.26 | 30.58 | 33.17 | |
| Qwen2.5-VL (7B)Type=Zero-Shot VLM, Frames=1282026.03 | 34.58 | 39.03 | 32.16 | 32.37 | |
| Qwen2.5-VLType=Fine-Tuned Model, Frames=1282026.03 | 34.55 | 38.8 | 31.88 | 32.38 | |
| VideoChat-Flash (7B)Type=Zero-Shot Long-Form VLM, Frames=30002026.03 | 32.93 | 36.29 | 26.83 | 29.63 | |
| LLaVA-OneVision (7B)Type=Zero-Shot VLM, Frames=322026.03 | 31.43 | 35.17 | 26.22 | 27.38 | |
| InternVL2.5 (7B)Type=Zero-Shot VLM, Frames=642026.03 | 31.2 | 34.98 | 31.78 | 29 | |
| MovieChat (7B)Type=Zero-Shot Long-Form VLM, Frames=20482026.03 | 31.19 | 35.32 | 25.01 | 28.45 | |
| InternVL2 (7B)Type=Zero-Shot VLM, Frames=322026.03 | 29.65 | 32.36 | 28.03 | 26.22 | |
| VideoChatGPT (7B)Type=Zero-Shot VLM, Frames=1002026.03 | 27.66 | 31.47 | 23.86 | 21.47 | |
| Qwen2.5 Instruct (7B)Type=Zero-Shot VLM, Use Descr.=true2026.03 | 25.52 | 30.81 | 25.83 | 22.01 | |
| VideoLLaVA (7B)Type=Zero-Shot VLM, Frames=82026.03 | 15.79 | 17.84 | 14.9 | 11.55 |