Text-to-Video Generation on Wan2.1-14B-T2V-distilled (81 frames, 480P, 4steps, no cfg)
31.997CLIP-SCOREbaseline
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| baseline2026.04 | 31.997 | 0.9 | 64.15 | — | 1 | 1 | — | |
| Chorus(0)Cache capacity=cold-start (0)2026.04 | 31.5006 | 0.9005 | 55.91 | 44.51 | 1.15 | 1.44 | 42.6 | |
| Chorus(1k)Cache capacity=warm-start (1k)2026.04 | 31.3081 | 0.9006 | 52.09 | 44.15 | 1.23 | 1.45 | 58.9 | |
| NIRVANA(0)Cache capacity=cold-start (0)2026.04 | 31.3038 | 0.8993 | 56.75 | 46.37 | 1.13 | 1.38 | 42.6 | |
| NIRVANA(1k)Cache capacity=warm-start (1k)2026.04 | 31.0635 | 0.8991 | 53.79 | 46.93 | 1.19 | 1.37 | 58.9 |