Object Rotation on curated synthetic benchmark (test)
0.2861Rotation ErrorTALK2MOVE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TALK2MOVE2026.01 | 0.2861 | 29.55 | 68.75 | |
| Bagel2026.01 | 0.324 | 13.64 | 1.25 | |
| QwenImageEdit2026.01 | 0.4129 | 9.3 | 7.5 | |
| Flux-Kontext2026.01 | 0.4259 | 6.82 | 16.25 | |
| GPT-Image-12026.01 | 0.4293 | 2.33 | 6.25 |