Video Generation on VBench Wan2.1 1.3B extrapolation, 61x720x1280 (test)
65.49Aesthetic QualityFull Attention
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Full AttentionModel=Wan2.1 1.3B, Resolution=61x720x1280, Kernel=FullAttn, Training Setting=Fine-tuned2026.01 | 65.49 | 96.03 | 76.39 | 65.15 | 93.64 | — | 758.9 | 246.8 | |
| VSAModel=Wan2.1 1.3B, Resolution=61x720x1280, Kernel=VSA, Training Setting=Fine-tuned2026.01 | 65.07 | 95.86 | 63.89 | 65.69 | 92.07 | 90 | 208.6 | 125.8 | |
| VMonarchModel=Wan2.1 1.3B, Resolution=61x720x1280, Kernel=VMonarch, Training Setting=Fine-tuned2026.01 | 64.39 | 96.13 | 69.44 | 65.84 | 93.82 | 87.5 | 243.3 | 136.8 |