Long-video Understanding on LVBench 1.0 (test)
48.4Overall ScoreVideo-XL-2
Evaluation Results
| Method | Links | |
|---|---|---|
| Video-XL-2Venue=Arxiv’25, LLM Size=8B, Frames=1fps2025.08 | 48.4 | |
| Video-EM (Qwen2.5-VL)LLM Size=7B, Frames=Avg 272025.08 | 45.7 | |
| Video-EM (Qwen2-VL)Venue=Arxiv’25, LLM Size=7B, Frames=Avg 272025.08 | 45.2 | |
| Qwen2-VLVenue=ArXiv’24, LLM Size=72B, Frames=1fps2025.08 | 44 | |
| VAMBAVenue=CVPR’25, LLM Size=10B, Frames=10242025.08 | 42.1 | |
| LLaVA-VideoVenue=ArXiv’24, LLM Size=7B, Frames=1802025.08 | 41.5 | |
| Qwen2-VL‡Venue=ArXiv’24, LLM Size=7B, Frames=642025.08 | 39.7 | |
| Qwen2.5-VL‡Venue=ArXiv’25, LLM Size=7B, Frames=642025.08 | 38.7 | |
| Qwen2-VL‡Venue=ArXiv’24, LLM Size=7B, Frames=322025.08 | 38.4 | |
| Video-XLVenue=CVPR’25, LLM Size=7B2025.08 | 36.8 | |
| Qwen2.5-VL‡Venue=ArXiv’25, LLM Size=7B, Frames=322025.08 | 36.6 | |
| GPT-4o2025.08 | 34.7 | |
| Gemini 1.5 pro2025.08 | 33.1 |