Benchmarking World Modeling in Dynamically Changing Environments
| Model | VisQual ↑ | MotSmooth ↑ | ObjConsist ↑ | 3DConsist ↑ | ORS ↑ | PSNR ↑ | SSIM ↑ | LPIPS ↓ | CamCtrl ↑ | ImgReward ↑ |
|---|---|---|---|---|---|---|---|---|---|---|
| CI2V Models | ||||||||||
| LingBot-World | 47.4 | 57.6 | 59.0 | 88.2 | 0.381 | 14.41 | 0.490 | 0.482 | 37.4 | 36.7 |
| Wan2.2 | 40.0 | 54.0 | 50.7 | 84.5 | 0.328 | 13.76 | 0.469 | 0.529 | 29.8 | 26.1 |
| FantasyWorld | 51.0 | 55.2 | 47.6 | 88.7 | 0.276 | 13.23 | 0.427 | 0.571 | 27.2 | 30.7 |
| HunyuanWorldPlay | 43.5 | 66.6 | 61.5 | 90.6 | 0.582 | 14.35 | 0.471 | 0.505 | 69.9 | 24.5 |
| HunyuanGameCraft | 54.2 | 51.9 | 46.6 | 85.9 | 0.266 | 12.81 | 0.388 | 0.603 | 54.2 | 8.9 |
| 3D-based Models | ||||||||||
| Matrix-Game 2.0 | 61.2 | 83.6 | 46.5 | 93.7 | 0.157 | 13.49 | 0.376 | 0.550 | 17.3 | 22.3 |
| Stable Virtual Camera | 43.3 | 63.1 | 59.5 | 88.5 | 0.294 | 15.36 | 0.523 | 0.455 | 65.2 | 22.3 |
| I2V Models | ||||||||||
| Open-SoRA | 49.7 | 68.3 | 47.2 | 89.7 | 0.182 | 12.54 | 0.384 | 0.566 | 16.8 | 31.3 |
| LTX-Video | 44.9 | 84.4 | 81.6 | 94.1 | 0.330 | 13.42 | 0.455 | 0.518 | 17.1 | 37.1 |
| CogVideoX | 40.1 | 59.8 | 54.0 | 94.0 | 0.251 | 12.07 | 0.480 | 0.592 | 12.0 | 34.9 |
💡 Notice: We are actively benchmarking more models. New entries will be added continuously!