Text-to-Image Generation on MS-COCO (30K)
1.15FID (30K)OrthoEraser
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OrthoEraserBase Model=Stable Diffusion 1.42026.03 | 1.15 | 31.33 | — | |
| Janus-ProSize=7.0B, Type=Uni., Strategy=Decoupled Enc., Dec.=Frozen2026.05 | 6.85 | 32 | — | |
| ImagenNumber of Parameters=3.0B2023.12 | 7.27 | 27 | — | |
| MUSE-3BNumber of Parameters=3.0B2023.12 | 7.88 | 32 | — | |
| Parti-3BNumber of Parameters=3.0B2023.12 | 8.1 | — | — | |
| GigaGANNumber of Parameters=1.0B2023.12 | 9.09 | 32.2 | — | |
| VA-πSize=775M, Type=AR, Strategy=Variational Align., Dec.=Frozen2026.05 | 9.23 | 29.1 | — | |
| DALL-E-2#Params=5.2B, #Images=250M2025.04 | 10.39 | — | — | |
| Diffusion-DPOSize=2.6B, Type=Diff., Strategy=DPO, Dec.=Frozen2026.05 | 10.5 | 31.5 | — | |
| Flow-GRPOSize=2.0B, Type=Flow, Strategy=Online GRPO, Dec.=Frozen2026.05 | 10.5 | — | — | |
| Parti-750MNumber of Parameters=0.8B2023.12 | 10.71 | — | — | |
| Emu3Size=7.0B, Type=AR, Strategy=Unified AR, Dec.=Frozen2026.05 | 11.6 | 28.6 | — | |
| SDXL-Base-1.0#Params=3.5B2025.04 | 12.15 | 31.99 | 35.12 | |
| GLIDE#Params=3.5B, #Images=250M2025.04 | 12.24 | — | — | |
| LDM-KL-8-G#Params=1.45B, #Images=400M2025.04 | 12.63 | — | 30.29 | |
| Small SD#Params=0.76B, #Images=229M2025.04 | 12.76 | 28.51 | 32.33 | |
| SD v1.4Number of Parameters=0.9B2023.12 | 12.94 | 32.5 | — | |
| SDM-v1.4#Params=1.04B, #Images=>2000M2025.04 | 13.05 | 29.58 | 36.76 | |
| C-Micro#Params=0.40B, #Images=02025.04 | 13.42 | 28.13 | 32.64 | |
| SnapFusion#Params=0.99B, #Images=>100M2025.04 | 13.6 | 29.5 | — | |
| C-Tiny#Params=0.49B, #Images=02025.04 | 13.9 | 28.6 | 33.18 | |
| GenTron-XL/2Number of Parameters=0.9B2023.12 | 14.21 | 32.6 | — | |
| C-Small#Params=0.61B, #Images=02025.04 | 14.43 | 28.88 | 34.58 | |
| C-Base#Params=0.73B, #Images=02025.04 | 14.45 | 29.04 | 34.92 | |
| B-Base#Params=0.76B, #Images=02025.04 | 14.47 | 29.32 | 36.5 | |
| GenTron-G/2Number of Parameters=3.1B2023.12 | 14.53 | 33.5 | — | |
| RankESize=775M, Type=AR, Strategy=Co-Evolved, Dec.=Unfrozen2026.05 | 15.21 | 33.76 | — | |
| LlamaGenSize=775M, Type=AR, Strategy=Pretrain, Dec.=Unfrozen2026.05 | 15.24 | 31.54 | — | |
| BK-SDM-Base#Params=0.76B, #Images=0.22M2025.04 | 15.76 | 28.78 | 33.79 | |
| B-Small#Params=0.66B, #Images=02025.04 | 16.22 | 28.04 | 35.99 | |
| SD 3.5 Medium#Params=7.9B2025.04 | 16.23 | 32.46 | 39.81 | |
| LlamaGen-SFTSize=775M, Type=AR, Strategy=SFT, Dec.=Frozen2026.05 | 16.58 | 31.86 | — | |
| SPMBase Model=Stable Diffusion 1.42026.03 | 16.64 | 31.01 | — | |
| SNCEBase Model=Stable Diffusion 1.42026.03 | 16.64 | 30.87 | — | |
| B-Tiny#Params=0.50B, #Images=02025.04 | 16.71 | 27.82 | 35.46 | |
| ESDBase Model=Stable Diffusion 1.42026.03 | 16.88 | 30.9 | — | |
| TransfusionSize=7.0B, Type=Unified, Strategy=Mixed-Modal, Dec.=Frozen2026.05 | 16.9 | 25.5 | — | |
| BK-SDM-Small#Params=0.66B, #Images=0.22M2025.04 | 16.98 | 26.77 | 31.68 | |
| BK-SDM-Tiny#Params=0.50B, #Images=0.22M2025.04 | 17.12 | 26.53 | 30.09 | |
| LlamaGen-RLSize=775M, Type=AR, Strategy=GRPO, Dec.=Frozen2026.05 | 17.76 | 32.45 | — | |
| SDXLNumber of Parameters=2.6B2023.12 | 17.82 | 32.9 | — | |
| RECEBase Model=Stable Diffusion 1.42026.03 | 18.25 | 30.95 | — | |
| SLD-MedBase Model=Stable Diffusion 1.42026.03 | 19.53 | 30.65 | — | |
| ECADModel=PixArt-α, Caching=Ours, Setting=fastest2025.06 | 19.54 | 31.28 | — | |
| ECADModel=PixArt-α, Caching=Ours, Setting=fast2025.06 | 20.58 | 31.4 | — | |
| CABase Model=Stable Diffusion 1.42026.03 | 21.55 | 31.21 | — | |
| ECADModel=FLUX.1-dev, Caching=Ours, Setting=fast2025.06 | 21.61 | 31.58 | — | |
| ECADModel=PixArt-α, Caching=Ours, Setting=faster2025.06 | 21.93 | 31.1 | — | |
| ToCaModel=PixArt-α, Caching=ToCa, Setting=N = 3, R = 60%2025.06 | 22.05 | 30.99 | — | |
| ECADModel=PixArt-Σ, Caching=Ours, Setting=fast2025.06 | 22.17 | 31.25 | — | |
| Würstchen-v2#Params=3.1B, #Images=1700M2025.04 | 22.4 | 26.76 | 32.87 | |
| UCEBase Model=Stable Diffusion 1.42026.03 | 22.87 | 29.92 | — | |
| DuCaModel=PixArt-α, Caching=DuCa, Setting=N = 3, R = 60%2025.06 | 23.13 | 31.03 | — | |
| Pixart-alpha#Params=5.4B, #Images=25M2025.04 | 23.43 | 30.72 | 34.54 | |
| FORAModel=FLUX.1-dev, Caching=FORA, Setting=N = 32025.06 | 23.51 | 31.3 | — | |
| ToCaModel=FLUX.1-dev, Caching=ToCa, Setting=N = 4, R = 90%2025.06 | 23.78 | 31.26 | — | |
| TGATEModel=PixArt-α, Caching=TGATE, Setting=m = 15, k = 12025.06 | 23.9 | 31.12 | — | |
| ToCaModel=PixArt-α, Caching=ToCa, Setting=N = 3, R = 90%2025.06 | 24.01 | 30.92 | — | |
| FORAModel=PixArt-α, Caching=FORA, Setting=N = 32025.06 | 24.5 | 31.35 | — | |
| NoneModel=PixArt-Σ, Caching=None2025.06 | 24.63 | 31.11 | — | |
| DuCaModel=PixArt-α, Caching=DuCa, Setting=N = 3, R = 90%2025.06 | 24.69 | 30.96 | — | |
| FORAModel=PixArt-α, Caching=FORA, Setting=N = 22025.06 | 24.8 | 31.37 | — | |
| NoneModel=PixArt-α, Caching=None2025.06 | 24.84 | 31.29 | — | |
| NoneModel=FLUX.1-dev, Caching=None2025.06 | 25.76 | 30.95 | — | |
| DiCacheModel=FLUX.1-dev, Caching=DiCache2025.06 | 26.18 | 31.12 | — | |
| ECADModel=FLUX.1-dev, Caching=Ours, Setting=fastest2025.06 | 26.66 | 31.63 | — | |
| FORAModel=PixArt-Σ, Caching=FORA, Setting=N = 32025.06 | 27.69 | 31.16 | — | |
| TaylorSeerModel=FLUX.1-dev, Caching=TaylorSeer, Setting=N = 5, O = 22025.06 | 29.66 | 30.19 | — | |
| TGATEModel=PixArt-α, Caching=TGATE, Setting=m = 10, k = 52025.06 | 29.78 | 28.29 | — | |
| TaylorSeerModel=FLUX.1-dev, Caching=TaylorSeer, Setting=N = 6, O = 12025.06 | 49.02 | 29.02 | — | |
| ToCaModel=PixArt-Σ, Caching=ToCa†, Setting=N = 3, R = 60%2025.06 | 54.8 | 30.34 | — | |
| ToCaModel=PixArt-Σ, Caching=ToCa†, Setting=N = 3, R = 90%2025.06 | 56.48 | 30.25 | — | |
| DuMoBase Model=Stable Diffusion 1.42026.03 | — | 30.87 | — | |
| SD1.4Base Model=Stable Diffusion 1.42026.03 | — | 31.34 | — | |
| SDXLSize=2.6B, Type=Diff., Strategy=Pretrain, Dec.=Frozen2026.05 | — | 31 | — |