Video Understanding on MMMU Video
55.91AccuracyHybrid-Frame Strategy
Evaluation Results
| Method | Links | |
|---|---|---|
| Hybrid-Frame StrategyBackbone=Qwen2.5-VL-7B, Frames=64F2026.03 | 55.91 | |
| Hybrid-Frame StrategyBackbone=Qwen2.5-VL-7B, Frames=32F2026.03 | 55.12 | |
| Hybrid-Frame StrategyBackbone=Qwen2.5-VL-7B, Frames=16F2026.03 | 51.78 | |
| Video-SFTBackbone=Qwen2.5-VL-7B, Frames=32F2026.03 | 51.44 | |
| Video-SFTBackbone=Qwen2.5-VL-7B, Frames=64F2026.03 | 51.44 | |
| Video-SFTBackbone=Qwen2.5-VL-7B, Frames=16F2026.03 | 50.44 | |
| Video-SFTBackbone=Qwen2.5-VL-7B, Frames=8F2026.03 | 47.56 | |
| Hybrid-Frame StrategyBackbone=Qwen2.5-VL-7B, Frames=8F2026.03 | 47.22 | |
| Hybrid-Frame StrategyBackbone=LLaVA-1.5-7B, Note=Evaluated with a fixed 8-frame inference2026.03 | 18.71 | |
| Video-SFTBackbone=LLaVA-1.5-7B, Frames=8F2026.03 | 18.56 | |
| Video-SFTBackbone=LLaVA-1.5-7B, Frames=16F2026.03 | 16.44 | |
| Video-SFTBackbone=LLaVA-1.5-7B, Frames=32F2026.03 | 12.92 |