Video Generation on VBench & VisionReward HunyuanVideo 241f
0.131VisionRewardDynamicRad (Dynamic)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DynamicRad (Dynamic)Model=HunyuanVideo, #Frames=241, Resolution=1280 x 768, Hardware=8xH100 cluster, Fine-tuned backbone=true, Training-free=true2026.04 | 0.131 | 96.9 | 65.9 | 68.7 | 83.6 | 378 | 2.17 | |
| Radial AttentionModel=HunyuanVideo, #Frames=241, Resolution=1280 x 768, Hardware=8xH100 cluster, Fine-tuned backbone=true2026.04 | 0.125 | 96.1 | 62.6 | 65.9 | 80.6 | 351 | 2.34 | |
| DynamicRad (Static)Model=HunyuanVideo, #Frames=241, Resolution=1280 x 768, Hardware=8xH100 cluster, Fine-tuned backbone=true, Training-free=true2026.04 | 0.118 | 95.1 | 61.2 | 66.8 | 86.8 | 324 | 2.54 | |
| Original (LoRA)Model=HunyuanVideo, #Frames=241, Resolution=1280 x 768, Hardware=8xH100 cluster, Fine-tuned backbone=true, LoRA=true2026.04 | 0.116 | 94.8 | 59.2 | 59.7 | 0 | 824 | 1 |