Text-to-Image Generation on COCO 10k-sample 2017
90Precision (P)Teacher model
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Teacher model2025.12 | 90 | 64 | 88 | 91 | 57.938 | 0.6575 | |
| Progressive Adversarial Distillation2025.12 | 89 | 63 | 98.8 | 92 | 65.9475 | 0.6525 | |
| ReDiF(GRPO)Algorithm=GRPO2025.12 | 85 | 56 | 102 | 88 | 63.53 | 0.6494 | |
| ReDiF(PPO with renyi)Algorithm=PPO, Divergence Function=renyi2025.12 | 85 | 70 | 96.6 | 96 | 63.287 | 0.6503 | |
| Progressive Distillation2025.12 | 84 | 53 | 105.2 | 89 | 65.7075 | 0.6504 | |
| Consistency Distillation2025.12 | 83 | 60 | 101 | 92 | 67.4574 | 0.6496 | |
| DDIM2025.12 | 81 | 57 | 101.8 | 93 | 68.5488 | 0.65 | |
| Distribution Matching Distillation2025.12 | 81 | 55 | 91.4 | 86 | 66.0154 | 0.6507 |