Video Understanding on LVbench (MAT, Wall-time speedup)
12.12MATHIPPO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HIPPOBackbone Model=Video-SALMONN2+ 7&72B2026.01 | 12.12 | 2.64 | |
| HIPPOBackbone Model=Qwen2.5-VL 7&72B2026.01 | 9.25 | 2.44 | |
| HIPPOBackbone Model=LLaVA-OneVision 7&72B2026.01 | 8.39 | 3.16 | |
| HIPPOBackbone Model=Qwen3-VL 7&32B2026.01 | 5.39 | 2.32 | |
| SD-treeBackbone Model=Qwen3-VL 7&32B2026.01 | 4.51 | 1.88 | |
| SpecVLMBackbone Model=Qwen3-VL 7&32B2026.01 | 4.2 | 2.11 | |
| Vanilla SDBackbone Model=Qwen3-VL 7&32B2026.01 | 3.88 | 1.39 | |
| SD-treeBackbone Model=Qwen2.5-VL 7&72B2026.01 | 3.53 | 1.83 | |
| SpecVLMBackbone Model=Qwen2.5-VL 7&72B2026.01 | 3.49 | 2.1 | |
| SpecVLMBackbone Model=Video-SALMONN2+ 7&72B2026.01 | 3.42 | 2.28 | |
| SpecVLMBackbone Model=LLaVA-OneVision 7&72B2026.01 | 3.32 | 2.37 | |
| SD-treeBackbone Model=Video-SALMONN2+ 7&72B2026.01 | 3.29 | 1.72 | |
| SD-treeBackbone Model=LLaVA-OneVision 7&72B2026.01 | 3.13 | 1.93 | |
| Vanilla SDBackbone Model=Qwen2.5-VL 7&72B2026.01 | 3.11 | 1.57 | |
| Vanilla SDBackbone Model=LLaVA-OneVision 7&72B2026.01 | 2.94 | 1.68 | |
| Vanilla SDBackbone Model=Video-SALMONN2+ 7&72B2026.01 | 2.89 | 1.57 |