Long-video understanding on VideoMME Long (3~120 min)
72.1ScoreGPT-4o
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4oFrames=1fps2026.06 | 72.1 | |
| Video-RAGSize=7B2026.06 | 59.8 | |
| LongVUSize=7B, Frames=1fps2026.06 | 59.5 | |
| LongVITA-128KSize=14B, Frames=2562026.06 | 58.8 | |
| VILAMPSize=7B, Frames=1fps2026.06 | 57.8 | |
| GPT-4VFrames=1fps2026.06 | 56.9 | |
| VideoChat-FlashSize=7B, Frames=5122026.06 | 55.4 | |
| LongVILA-R1Size=7B, Frames=512, Context=132k2026.06 | 55.2 | |
| Hour-LLaVASize=7B, Frames=1fps2026.06 | 55 | |
| CoVERSize=7B, Frames=768L + 32H, Context=16k2026.06 | 54.7 | |
| VITALSize=7B, Frames=1024, Context=32k2026.06 | 54 | |
| Qwen2.5-VL*Size=7B, Frames=128, Context=16k2026.06 | 53.6 | |
| LongVILASize=7B, Frames=2562026.06 | 53 | |
| Video-MTRSize=7B, Frames=32, Context=4k2026.06 | 51 | |
| VideoMindSize=7B2026.06 | 49.2 | |
| LongVASize=7B, Frames=1282026.06 | 47.6 | |
| LLaVA-OneVisionSize=7B, Frames=322026.06 | 46.7 | |
| VideoLLaMA2Size=7B, Frames=322026.06 | 43.8 | |
| VideoChat-TSize=7B, Frames=122026.06 | 41.9 | |
| VideoChat2Size=7B, Frames=162026.06 | 39.2 | |
| Video-LLaVASize=7B, Frames=82026.06 | 38.1 | |
| ShareGPT4VideoSize=8B, Frames=162026.06 | 37.9 |