Interactive Video Generation (Hard Cut) on VBench Long
34.27Text AlignmentLongLive + Echo-Forcing
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LongLive + Echo-ForcingFine-tuning protocol=Fine-tuned2026.05 | 34.27 | 83.39 | 57.75 | 55.01 | 69.77 | |
| Echo-ForcingFine-tuning protocol=Unfine-tuned2026.05 | 33.67 | 79.39 | 89.62 | 58 | 69.64 | |
| ∞-RopeFine-tuning protocol=Unfine-tuned2026.05 | 32.63 | 78.88 | 85.89 | 57.12 | 67.79 | |
| LongLive + FlushFine-tuning protocol=Fine-tuned2026.05 | 31.97 | 81.99 | 56.33 | 54.49 | 69.01 | |
| LongLiveFine-tuning protocol=Fine-tuned2026.05 | 30.59 | 81.53 | 57.78 | 54.46 | 67.76 | |
| Self-Forcing + RecacheFine-tuning protocol=Unfine-tuned2026.05 | 29.88 | 76.43 | 86.77 | 50.02 | 50.17 |