Physically Plausible Video Generation on Physics-IQ 26
53.1Solid Mechanics ScorePhyCo (ControlNet + VLM Loss)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PhyCo (ControlNet + VLM Loss)Generation Setting=Train-time conditions: 57 frames @ 24FPS + last-frame repetition2026.04 | 53.1 | 44.3 | 20.3 | 20.8 | 35.9 | 43.6 | |
| PhyCo (ControlNet)Generation Setting=Train-time conditions: 57 frames @ 24FPS + last-frame repetition2026.04 | 49.7 | 37.8 | 16.3 | 19.9 | 18.2 | 38.9 | |
| PhyCo (ControlNet + VLM Loss)Generation Setting=Test time extrapolated generation: 120 frames @ 24FPS2026.04 | 44.1 | 31.2 | 20.1 | 17.2 | 33.1 | 36.3 | |
| PhyCo (Text only)Generation Setting=Train-time conditions: 57 frames @ 24FPS + last-frame repetition2026.04 | 43.9 | 38.5 | 17.5 | 21.7 | 26.8 | 36.5 | |
| VLIPPGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 42.3 | 34.1 | 16.9 | 13.4 | 8.8 | 34.6 | |
| PhyCo (ControlNet)Generation Setting=Test time extrapolated generation: 120 frames @ 24FPS2026.04 | 42.3 | 30.7 | 19.3 | 12.6 | 40.1 | 35.3 | |
| PhyCo (Text only)Generation Setting=Test time extrapolated generation: 120 frames @ 24FPS2026.04 | 36.5 | 28.9 | 18.9 | 12.6 | 32 | 30.9 | |
| SG-I2VGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 34.6 | 31.2 | 15.9 | 13.1 | 8.4 | 29.7 | |
| Cosmos-Predict2-2BGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 31.7 | 25.2 | 26.2 | 9.1 | 16.9 | 27.7 | |
| Cogvideo-I2V-5BGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 30.4 | 29.8 | 16.7 | 13.3 | 8.5 | 27.1 | |
| LTX-Video-I2VGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 30.2 | 29.8 | 15.9 | 13.2 | 8.4 | 26.8 | |
| SVD-XTGeneration Setting=120 frames @ 24FPS (Extrapolated)2026.04 | 21.9 | 20.5 | 6.8 | 8.4 | 17.1 | 19.1 |