Long-video understanding on LongVideoBench 0~60 min
66.7ScoreGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oFrames=1fps2026.06 | 66.7 | |
| VILAMPSize=7B, Frames=1fps2026.06 | 61.2 | |
| LongVITA-128KSize=14B, Frames=2562026.06 | 60.9 | |
| Hour-LLaVASize=7B, Frames=1fps2026.06 | 60.4 | |
| Video-RAGSize=7B2026.06 | 58.7 | |
| CoVERSize=7B, Frames=768L + 32H, Context=16k2026.06 | 58.3 | |
| LongVILA-R1Size=7B, Frames=512, Context=132k2026.06 | 58 | |
| LongVILASize=7B, Frames=2562026.06 | 57.1 | |
| Video-XL-ProSize=7B, Frames=2402026.06 | 56.7 | |
| ConanSize=7B, Frames=322026.06 | 56.6 | |
| VideoMindSize=7B2026.06 | 56.3 | |
| Qwen2.5-VL*Size=7B, Frames=128, Context=16k2026.06 | 56 | |
| VambaSize=10B, Frames=10242026.06 | 55.9 | |
| Video-XLSize=7B, Frames=2562026.06 | 50.7 | |
| LLaVA-NeXT-VideoSize=7B, Frames=322026.06 | 43.5 | |
| ShareGPT4VideoSize=8B, Frames=162026.06 | 39.7 | |
| Video-LLaVASize=7B, Frames=82026.06 | 39.1 |