Physical Law Evaluation on PhyGenBench
62.5Mechanics AccuracyPILA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PILAsize=14B, variant=†2026.06 | 62.5 | 84 | 58.9 | 61.7 | 68.3 | |
| PILAsize=1.3B2026.06 | 57.5 | 71.3 | 57.8 | 55.8 | 61.5 | |
| Wan 2.1size=1.3B2026.06 | 48.3 | 64 | 32.2 | 34.2 | 46.7 | |
| CogVideoXsize=5B2026.06 | 48.3 | 67.3 | 43.3 | 42.5 | 51.9 | |
| Wan 2.2size=14B2026.06 | 45.8 | 75.3 | 44.4 | 41.7 | 53.8 | |
| WISAsize=14B2026.06 | 45.8 | 70 | 35.6 | 37.5 | 49.4 | |
| VideoREPAsize=2B2026.06 | 43.3 | 62.7 | 43.3 | 41.7 | 49 | |
| Hunyuansize=80G2026.06 | 37.5 | 42 | 21.1 | 32.5 | 34.6 |