Interactive Video Generation (Scene Recall) on VBench Long
32.58Text AlignmentEcho-Forcing
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Echo-ForcingFine-tuning protocol=Unfine-tuned2026.05 | 32.58 | 83.11 | 81.57 | 58.76 | 68.88 | |
| LongLive + Echo-ForcingFine-tuning protocol=Fine-tuned2026.05 | 32.58 | 83.48 | 87.57 | 54.98 | 69.13 | |
| LongLive + FlushFine-tuning protocol=Fine-tuned2026.05 | 30.18 | 82.77 | 86.23 | 53.77 | 67.43 | |
| ∞-RopeFine-tuning protocol=Unfine-tuned2026.05 | 29.47 | 79.31 | 78.39 | 56.31 | 66.73 | |
| Self-Forcing + RecacheFine-tuning protocol=Unfine-tuned2026.05 | 28.58 | 78.12 | 74.71 | 50.98 | 51.02 | |
| LongLiveFine-tuning protocol=Fine-tuned2026.05 | 28.56 | 82.23 | 86.5 | 53.31 | 66.32 |