Text-to-Video on 1,024 prompts (held-out)
4.81VQMapReduce LoRA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MapReduce LoRAReward=VQ+MQ2025.11 | 4.81 | 1.81 | |
| DanceGRPOReward=VQ2025.11 | 4.37 | 1.2 | |
| Rewarded SoupReward=VQ+MQ2025.11 | 4.13 | 1.43 | |
| DanceGRPOReward=MQ2025.11 | 3.8 | 1.66 | |
| HunyuanVideoReward=None2025.11 | 3.25 | 0.95 |