Physical Plausibility Evaluation on VideoPhy Hard 2
86.1PC ScoreVideoREPA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VideoREPABackbone=LTX-Video-2B2026.05 | 86.1 | 1.1 | 1.1 | |
| NewtonBackbone=LTX-Video-2B2026.05 | 78.89 | 12.78 | 12.22 | |
| WISABackbone=LTX-Video-2B2026.05 | 74.4 | 2.2 | 2.2 | |
| HunyuanVideo2026.05 | 63.33 | 8.89 | 6.67 | |
| LTX-Video-2B2026.05 | 63.33 | 6.67 | 4.44 | |
| ProprioRefinement Strategy=Search (BoN4) + refinement, Search=BoN4, Scoring Method=S_hybrid2026.05 | 55 | 31.1 | 21.1 | |
| CogVideoX-5B2026.05 | 54.44 | 9.44 | 5 | |
| ProprioRefinement Strategy=Search (BoN4) + refinement, Search=BoN4, Scoring Method=S_global2026.05 | 54.4 | 29.4 | 21.1 | |
| ProprioRefinement Strategy=Search (BoN8) + refinement, Search=BoN8, Scoring Method=S_motion2026.05 | 54.4 | 31.7 | 23.3 | |
| ProprioRefinement Strategy=Search (BoN4) + refinement, Search=BoN4, Scoring Method=S_motion2026.05 | 54.3 | 29.7 | 21.1 | |
| ProprioRefinement Strategy=Search (BoN8) + refinement, Search=BoN8, Scoring Method=S_hybrid2026.05 | 54.2 | 29.6 | 20.7 | |
| WMRewardRefinement Strategy=Search (BoN8) + refinement, Guidance=True, Search=BoN82026.05 | 52.2 | 30.6 | 21.1 | |
| ProprioRefinement Strategy=Refinement only, Scoring Method=S_hybrid2026.05 | 50.6 | 32.2 | 21.1 | |
| WMRewardRefinement Strategy=Search (BoN4) + refinement, Guidance=True, Search=BoN42026.05 | 48.9 | 32.8 | 19.4 | |
| ProprioRefinement Strategy=Refinement only, Scoring Method=S_motion2026.05 | 47.2 | 32.2 | 20.6 | |
| Cosmos-Predict2.52026.05 | 46.11 | 12.22 | 7.78 | |
| Turbo Wan2.22026.05 | 45.6 | 31.1 | 19.4 | |
| WMRewardRefinement Strategy=Refinement only, Guidance=True2026.05 | 45 | 32.8 | 21.7 | |
| ProprioRefinement Strategy=Refinement only, Scoring Method=S_global2026.05 | 43.9 | 29.4 | 20 | |
| Wan2.2-TI2V-5B2026.05 | 37.2 | 8.3 | 3.9 |