Text-to-Image Generation on GenEval2 In-Domain
92.6GenEval2 ScoreFlow-DPPO + CPS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Flow-DPPO + CPSBase Model=FLUX2-klein-base-9B, Strategy=single-reward RL fine-tuning2026.06 | 92.6 | 0.315 | 21.97 | 0.279 | |
| Flow-DPPO + CPSBase Model=FLUX.1-dev, Strategy=single-reward RL fine-tuning2026.06 | 91.6 | 0.331 | 23.29 | 0.322 | |
| Flow-CPSBase Model=FLUX.1-dev, Strategy=single-reward RL fine-tuning2026.06 | 91.2 | 0.328 | 23.2 | 0.317 | |
| Flow-DPPOBase Model=FLUX.1-dev, Strategy=single-reward RL fine-tuning2026.06 | 90.7 | 0.331 | 23.15 | 0.323 | |
| Flow-GRPOBase Model=FLUX.1-dev, Strategy=single-reward RL fine-tuning2026.06 | 87.8 | 0.331 | 23.03 | 0.311 | |
| GRPO-GuardBase Model=FLUX.1-dev, Strategy=single-reward RL fine-tuning2026.06 | 87.6 | 0.333 | 22.69 | 0.293 | |
| GRPO-GuardBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 85.8 | 0.328 | 22.03 | 0.252 | |
| Flow-DPPOBase Model=FLUX2-klein-base-9B, Strategy=single-reward RL fine-tuning2026.06 | 85.1 | 0.331 | 22.22 | 0.294 | |
| Flow-GRPOBase Model=FLUX2-klein-base-9B, Strategy=single-reward RL fine-tuning2026.06 | 84.5 | 0.314 | 21.82 | 0.276 | |
| Flow-DPPO + CPSBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 84.1 | 0.316 | 21.99 | 0.262 | |
| GRPO-GuardBase Model=FLUX2-klein-base-9B, Strategy=single-reward RL fine-tuning2026.06 | 82.8 | 0.312 | 20.52 | 0.21 | |
| Flow-CPSBase Model=FLUX2-klein-base-9B, Strategy=single-reward RL fine-tuning2026.06 | 82.7 | 0.311 | 21.82 | 0.261 | |
| Flow-DPPOBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 78.9 | 0.319 | 22.06 | 0.263 | |
| Flow-CPSBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 74.8 | 0.313 | 21.68 | 0.235 | |
| Diffusion-NFTBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 64.5 | 0.307 | 21.69 | 0.251 | |
| Flow-DPPOBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 57.7 | 36.4 | 25.76 | 41.8 | |
| Flow-DPPOBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 57.7 | 0.364 | 25.76 | 0.418 | |
| Flow-GRPOBase Model=SD3.5-medium, Strategy=single-reward RL fine-tuning2026.06 | 56.6 | 0.297 | 21.21 | 0.219 | |
| Flow-DPPO + CPSBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 55.2 | 38.6 | 26.15 | 42.7 | |
| Flow-DPPO + CPSBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 55.2 | 0.386 | 26.15 | 0.427 | |
| Flow-DPPO + CPSBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 51.6 | 36.9 | 25.72 | 41.5 | |
| Flow-DPPO + CPSBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 51.6 | 0.369 | 25.72 | 0.415 | |
| GRPO-GuardBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 49 | 37.5 | 25.27 | 41.1 | |
| GRPO-GuardBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 49 | 0.375 | 25.27 | 0.411 | |
| Flow-DPPOBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 48.1 | 34.5 | 25.63 | 40.9 | |
| Flow-DPPOBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 48.1 | 0.345 | 25.63 | 0.409 | |
| GRPO-GuardBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 47.8 | 35.3 | 25.64 | 40.9 | |
| GRPO-GuardBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 47.8 | 0.353 | 25.64 | 0.409 | |
| Diffusion-NFTBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 47.3 | 33.6 | 24.87 | 38.9 | |
| Diffusion-NFTBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 47.3 | 0.336 | 24.87 | 0.389 | |
| Flow-CPSBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 47.1 | 36.1 | 25.7 | 41.6 | |
| Flow-CPSBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 47.1 | 0.361 | 25.7 | 0.416 | |
| Flow-GRPOBase Model=FLUX2-klein-base-9B, Training Strategy=multi-reward RL fine-tuning2026.06 | 46.8 | 37.1 | 25.61 | 41.2 | |
| Flow-GRPOBase Model=FLUX2-klein-base-9B, Strategy=multi-reward RL fine-tuning2026.06 | 46.8 | 0.371 | 25.61 | 0.412 | |
| Flow-CPSBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 44.6 | 35.9 | 25.51 | 40.7 | |
| Gemini 2.5 Flash ImageBase Model=Gemini 2.5 Flash Image, Strategy=State-of-the-Art T2I Models2026.06 | 44.6 | — | — | — | |
| Flow-CPSBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 44.6 | 0.359 | 25.51 | 0.407 | |
| Diffusion-NFTBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 42.5 | 33.4 | 25.3 | 39.4 | |
| Diffusion-NFTBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 42.5 | 0.334 | 25.3 | 0.394 | |
| Flow-GRPOBase Model=SD3.5-medium, Training Strategy=multi-reward RL fine-tuning2026.06 | 39.9 | 35.8 | 25.09 | 39.9 | |
| Flow-GRPOBase Model=SD3.5-medium, Strategy=multi-reward RL fine-tuning2026.06 | 39.9 | 0.358 | 25.09 | 0.399 | |
| Qwen-ImageBase Model=Qwen-Image, Strategy=State-of-the-Art T2I Models2026.06 | 33.8 | — | — | — | |
| FLUX2-klein-base-9BBase Model=FLUX2-klein-base-9B, Training Strategy=Pretrained baselines (before RL)2026.06 | 25.4 | 28.1 | 20.92 | 22.8 | |
| FLUX2-klein-base-9BBase Model=FLUX2-klein-base-9B, Strategy=Pretrained (before RL)2026.06 | 25.4 | 0.281 | 20.92 | 0.228 | |
| FLUX.1-devBase Model=FLUX.1-dev, Training Strategy=Pretrained baselines (before RL)2026.06 | 23.3 | 29.7 | 23.26 | 31.5 | |
| FLUX.1-devBase Model=FLUX.1-dev, Strategy=Pretrained (before RL)2026.06 | 23.3 | 0.297 | 23.26 | 0.315 | |
| Bagel + CoTBase Model=Bagel + CoT, Strategy=State-of-the-Art T2I Models2026.06 | 23.1 | — | — | — | |
| SD3.5-largeBase Model=SD3.5-large, Strategy=State-of-the-Art T2I Models2026.06 | 22.8 | — | — | — | |
| SD3.5-mediumBase Model=SD3.5-medium, Training Strategy=Pretrained baselines (before RL)2026.06 | 12.4 | 25 | 21 | 21.3 | |
| SD3.5-mediumBase Model=SD3.5-medium, Strategy=Pretrained (before RL)2026.06 | 12.4 | 0.25 | 21 | 0.213 |