Reasoning-based Image Editing on RISE 43/52 (test)
34.1Temporal ScoreGPT-Image-1
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-Image-1Params=Not specified2026.02 | 34.1 | 32.2 | 37 | 10.6 | 28.9 | |
| Nano BananaParams=Not specified2026.02 | 25.9 | 47.8 | 37 | 18.8 | 32.8 | |
| DeepGen 1.0 (SFT)Params=3B + 2B, training_mode=SFT2026.02 | 15.3 | 18.9 | 14 | 4.7 | 13.3 | |
| Seedream 4.0Params=Not specified2026.02 | 12.9 | 12.2 | 11 | 7.1 | 10.8 | |
| DeepGen 1.0 (RL)Params=3B + 2B, training_mode=RL2026.02 | 12.9 | 14.4 | 13 | 2.4 | 10.8 | |
| BAGELParams=14B, CoT reasoning=true2026.02 | 5.9 | 17.8 | 21 | 1.2 | 11.9 | |
| Qwen-Image edit [2509]Params=7B + 20B2026.02 | 4.7 | 10 | 17 | 2.4 | 8.9 | |
| FLUX.1-Kontext [Dev]Params=12B2026.02 | 2.3 | 5.5 | 13 | 1.2 | 5.8 |