Video Understanding on VKnowU
67.7ScoreVideoKnow+
Evaluation Results
| Method | Links | |
|---|---|---|
| VideoKnow+Base Model=Qwen2.5-VL-7B-Instruct, Resolution=256x28x28, Max Frames=322025.11 | 67.7 | |
| Video-R1-7BResolution=256x28x28, Max Frames=322025.11 | 65.3 | |
| VideoRFT-7BResolution=256x28x28, Max Frames=322025.11 | 64.6 | |
| GRPO-ZeroBase Model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=GRPO-Zero, Resolution=256x28x28, Max Frames=322025.11 | 64.3 | |
| mPLUG-Owl3-8BResolution=256x28x28, Max Frames=322025.11 | 64.2 | |
| Zero-ShotBase Model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=Zero-Shot, Resolution=256x28x28, Max Frames=322025.11 | 64 | |
| See-Think-Answer SFTBase Model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=SFT, Resolution=256x28x28, Max Frames=322025.11 | 63.9 | |
| LLaVA-OneVision-7BResolution=256x28x28, Max Frames=322025.11 | 63.3 | |
| VideoLLaMA2Resolution=256x28x28, Max Frames=322025.11 | 55.7 |