Text-to-image generation on COCO (unspecified split)
5.74FIDSTRIDE
Evaluation Results
| Method | Links | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| STRIDEModel=FLUX.1-schnell, Sampling Steps=1-Step2026.05 | 5.74 | — | — | — | — | — | — | — | — | — | — | — | 0.683 | 3.268 | 27.66 | 0.283 | — | — | — | |
| BaselineModel=FLUX.1-schnell, Sampling Steps=1-Step2026.05 | 6.27 | — | — | — | — | — | — | — | — | — | — | — | 0.738 | 2.794 | 26.69 | 0.293 | — | — | — | |
| CADSModel=FLUX.1-schnell, Sampling Steps=1-Step2026.05 | 6.75 | — | — | — | — | — | — | — | — | — | — | — | 0.659 | 3.433 | 26.44 | 0.291 | — | — | — | |
| DDCModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 7.12 | — | — | — | — | — | — | — | — | — | — | — | 0.679 | 3.301 | 26.7 | 0.276 | — | — | — | |
| STRIDE No-PCAModel=FLUX.1-schnell, Sampling Steps=1-Step2026.05 | 8.09 | — | — | — | — | — | — | — | — | — | — | — | 0.762 | 2.603 | 28.05 | 0.276 | — | — | — | |
| STRIDEModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 8.68 | — | — | — | — | — | — | — | — | — | — | — | 0.725 | 2.905 | 26.69 | 0.289 | — | — | — | |
| Input NoiseModel=FLUX.1-schnell, Sampling Steps=1-Step2026.05 | 9.03 | — | — | — | — | — | — | — | — | — | — | — | 0.661 | 3.443 | 26.11 | 0.288 | — | — | — | |
| Input NoiseModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 9.14 | — | — | — | — | — | — | — | — | — | — | — | 0.749 | 2.703 | 26.72 | 0.291 | — | — | — | |
| STRIDE No-PCAModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 9.41 | — | — | — | — | — | — | — | — | — | — | — | 0.766 | 2.573 | 26.82 | 0.289 | — | — | — | |
| BaselineModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 9.62 | — | — | — | — | — | — | — | — | — | — | — | 0.771 | 2.527 | 26.77 | 0.29 | — | — | — | |
| CADSModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 17.54 | — | — | — | — | — | — | — | — | — | — | — | 0.391 | 5.888 | 22.56 | 0.263 | — | — | — | |
| Large Model2026.03 | 27.64 | 0.3181 | — | — | — | — | — | — | — | — | 20.72 | 1 | — | — | — | — | — | — | — | |
| SegQuant-GBackbone=SDXL-UNet, W/A=8/8(fp)2025.07 | 28 | — | — | — | 0.839 | 0.427 | 16.46 | 0.093 | 23.9 | 0.842 | — | — | — | — | — | — | — | — | — | |
| SegQuant-GBackbone=SDXL-UNet, W/A=8/8(int)2025.07 | 28.13 | — | — | — | 0.652 | 0.408 | 16.71 | 0.138 | 21.63 | 0.793 | — | — | — | — | — | — | — | — | — | |
| PTQ4DiTBackbone=SDXL-UNet, W/A=8/8(int)2025.07 | 28.15 | — | — | — | 0.639 | 0.407 | 16.7 | 0.213 | 19.17 | 0.725 | — | — | — | — | — | — | — | — | — | |
| Q-DiffusionBackbone=SDXL-UNet, W/A=8/8(fp)2025.07 | 28.24 | — | — | — | 0.843 | 0.428 | 16.46 | 0.104 | 23.32 | 0.832 | — | — | — | — | — | — | — | — | — | |
| SegQuant-ABackbone=SDXL-UNet, W/A=8/8(fp)2025.07 | 28.36 | — | — | — | 0.844 | 0.428 | 16.46 | 0.104 | 23.31 | 0.832 | — | — | — | — | — | — | — | — | — | |
| SegQuant-ABackbone=SDXL-UNet, W/A=8/8(int)2025.07 | 28.37 | — | — | — | 0.645 | 0.408 | 16.71 | 0.145 | 21.32 | 0.785 | — | — | — | — | — | — | — | — | — | |
| BaselineBackbone=SDXL-UNet, W/A=FP162025.07 | 28.44 | — | — | — | 0.841 | 0.429 | 16.54 | — | — | — | — | — | — | — | — | — | — | — | — | |
| HybridStitchmask ratio=20 %2026.03 | 30.34 | 0.3181 | — | — | — | — | — | 0.37 | — | — | 12.21 | 1.7 | — | — | — | — | — | — | — | |
| HybridStitchmask ratio=30 %2026.03 | 30.43 | 0.3184 | — | — | — | — | — | 0.42 | — | — | 11.31 | 1.83 | — | — | — | — | — | — | — | |
| GSPNResolution=1024 x 10242026.05 | 30.86 | 0.307 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HybridStitchmask ratio=40 %2026.03 | 30.96 | 0.3183 | — | — | — | — | — | 0.38 | — | — | 12.01 | 1.72 | — | — | — | — | — | — | — | |
| HybridStitchmask ratio=10 %2026.03 | 31.03 | 0.3182 | — | — | — | — | — | 0.35 | — | — | 12.03 | 1.72 | — | — | — | — | — | — | — | |
| SRDiffusion2026.03 | 31.67 | 0.3189 | — | — | — | — | — | 0.69 | — | — | 13.37 | 1.55 | — | — | — | — | — | — | — | |
| PTQ4DiTBackbone=SD3.5-DiT, W/A=8/8(int)2025.07 | 31.79 | — | — | — | 0.912 | 0.458 | 16.37 | 0.403 | 12.69 | 0.593 | — | — | — | — | — | — | — | — | — | |
| T-Stitch2026.03 | 31.87 | 0.3201 | — | — | — | — | — | 0.72 | — | — | 14.74 | 1.41 | — | — | — | — | — | — | — | |
| SegQuant-GBackbone=SD3.5-DiT, W/A=8/8(int)2025.07 | 32.02 | — | — | — | 0.991 | 0.457 | 16.36 | 0.376 | 13.14 | 0.601 | — | — | — | — | — | — | — | — | — | |
| TACBackbone=SD3.5-DiT, W/A=8/8(int)2025.07 | 32.22 | — | — | — | 0.863 | 0.458 | 16.31 | 0.417 | 12.53 | 0.583 | — | — | — | — | — | — | — | — | — | |
| SegQuant-ABackbone=SD3.5-DiT, W/A=8/8(int)2025.07 | 32.45 | — | — | — | 1.02 | 0.467 | 16.35 | 0.362 | 13.27 | 0.618 | — | — | — | — | — | — | — | — | — | |
| SD-v1.5Resolution=1024 x 1024, Variant=baseline2026.05 | 32.71 | 0.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLUX-1.DevSpeed (secs/image)=12, Inference steps=502026.02 | 33.07 | 0.314 | 0.3156 | 1.0291 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Fast GSPNResolution=1024 x 10242026.05 | 33.21 | 0.286 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLUX-1.DevSpeed (secs/image)=6.72, Inference steps=282026.02 | 33.35 | 0.312 | 0.314 | 1.0107 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDiTSpeed (secs/image)=5.52026.02 | 33.42 | 0.317 | 0.3136 | 1.0284 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDiT + TeacacheSpeed (secs/image)=3.4, delta=0.42026.02 | 33.6 | 0.315 | 0.3117 | 1.0182 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTQDBackbone=SD3.5-DiT, W/A=8/8(int)2025.07 | 33.78 | — | — | — | 0.822 | 0.434 | 16.47 | 0.454 | 10.79 | 0.523 | — | — | — | — | — | — | — | — | — | |
| FLUX-1.DevSpeed (secs/image)=3.6, Inference steps=152026.02 | 34.02 | 0.311 | 0.3121 | 0.9865 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineBackbone=SD3.5-DiT, W/A=FP162025.07 | 34.7 | — | — | — | 1.027 | 0.476 | 16.86 | — | — | — | — | — | — | — | — | — | — | — | — | |
| TaylorSeerSpeed (secs/image)=6, N=3, O=22026.02 | 34.74 | 0.303 | 0.3085 | 0.9721 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TeaCacheSpeed (secs/image)=6, delta=0.62026.02 | 34.95 | 0.303 | 0.3071 | 0.9968 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TaylorSeerSpeed (secs/image)=3.5, N=6, O=12026.02 | 35.02 | 0.302 | 0.304 | 0.9535 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TeaCacheSpeed (secs/image)=5.33, delta=0.82026.02 | 35.57 | 0.303 | 0.3075 | 0.978 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SegQuant-GBackbone=FLUX-DiT, W/A=8/8(int)2025.07 | 36.02 | — | — | — | 0.9 | 0.444 | 16.54 | 0.143 | 19.94 | 0.784 | — | — | — | — | — | — | — | — | — | |
| Q-DiffusionBackbone=FLUX-DiT, W/A=8/8(int)2025.07 | 36.04 | — | — | — | 0.883 | 0.446 | 16.52 | 0.302 | 15.23 | 0.624 | — | — | — | — | — | — | — | — | — | |
| BaselineBackbone=FLUX-DiT, W/A=BF162025.07 | 36.25 | — | — | — | 0.914 | 0.445 | 16.89 | — | — | — | — | — | — | — | — | — | — | — | — | |
| LinfusionResolution=1024 x 1024, Normalization=w/ norm2026.05 | 36.33 | 0.285 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SegQuant-ABackbone=FLUX-DiT, W/A=8/8(int)2025.07 | 36.48 | — | — | — | 0.907 | 0.445 | 16.52 | 0.155 | 19.4 | 0.77 | — | — | — | — | — | — | — | — | — | |
| SegQuant-GBackbone=FLUX-DiT, W/A=4/8(int)2025.07 | 36.88 | — | — | — | 0.882 | 0.439 | 16.56 | 0.232 | 17.03 | 0.7 | — | — | — | — | — | — | — | — | — | |
| SVDQuantBackbone=FLUX-DiT, W/A=4/8(int)2025.07 | 36.98 | — | — | — | 0.88 | 0.436 | 16.57 | 0.242 | 16.85 | 0.693 | — | — | — | — | — | — | — | — | — | |
| Mamba2Resolution=1024 x 1024, Normalization=w/ norm2026.05 | 37.02 | 0.273 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CRGBackbone=SD v1.4, Sampler=50-step DDPM, Negative Prompt=Bleeding, Blood, Gun, Weapon, Wounded2026.06 | 37.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.84 | — | — | — | — | |
| ESDConflict-Aware=false, Model modification required=true2026.02 | 38.15 | 0.3135 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTQ4DiTBackbone=FLUX-DiT, W/A=8/8(int)2025.07 | 38.82 | — | — | — | 0.716 | 0.46 | 16.45 | 0.328 | 14.73 | 0.62 | — | — | — | — | — | — | — | — | — | |
| SVDQuantBackbone=SD3.5-DiT, W/A=4/8(int)2025.07 | 39.22 | — | — | — | 0.855 | 0.454 | 16.4 | 0.432 | 12.32 | 0.582 | — | — | — | — | — | — | — | — | — | |
| ESDConflict-Aware=false2026.02 | 39.38 | 0.3139 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SPELLModel=SD3.5 Turbo, Sampling Steps=4-Step2026.05 | 39.91 | — | — | — | — | — | — | — | — | — | — | — | 0.603 | 3.971 | 24.44 | 0.247 | — | — | — | |
| SegQuant-GBackbone=SD3.5-DiT, W/A=4/8(int)2025.07 | 40.7 | — | — | — | 0.843 | 0.438 | 16.45 | 0.434 | 12.4 | 0.578 | — | — | — | — | — | — | — | — | — | |
| PTQ4DiTBackbone=FLUX-DiT, W/A=4/8(int)2025.07 | 41.16 | — | — | — | 0.096 | 0.417 | 16.65 | 0.561 | 11.63 | 0.537 | — | — | — | — | — | — | — | — | — | |
| SAFREEConflict-Aware=false, Model modification required=false2026.02 | 43.78 | 0.3053 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAFREEConflict-Aware=false2026.02 | 45.38 | 0.3051 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CASG+SAFREEConflict-Aware=true, Model modification required=false2026.02 | 46.25 | 0.3035 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CASG+SAFREEConflict-Aware=true2026.02 | 48.06 | 0.3025 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SafetyDPOConflict-Aware=false, Model modification required=true2026.02 | 49.64 | 0.3061 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MambaResolution=1024 x 1024, Normalization=w/ norm2026.05 | 50.3 | 0.263 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAFREEBackbone=SD v1.4, Sampler=50-step DDPM, Negative Prompt=Bleeding, Blood, Gun, Weapon, Wounded2026.06 | 50.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.42 | — | — | — | — | |
| NPBackbone=SD v1.4, Sampler=50-step DDPM, Negative Prompt=Bleeding, Blood, Gun, Weapon, Wounded2026.06 | 50.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.41 | — | — | — | — | |
| SafetyDPOConflict-Aware=false2026.02 | 51.2 | 0.3059 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CASG+SLDConflict-Aware=true, Model modification required=false2026.02 | 52 | 0.2936 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SLDConflict-Aware=false, Model modification required=false2026.02 | 52.11 | 0.2922 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CASG+SLDConflict-Aware=true2026.02 | 52.29 | 0.2934 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SLDConflict-Aware=false2026.02 | 53.32 | 0.2928 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DNGBackbone=SD v1.4, Sampler=50-step DDPM, Negative Prompt=Bleeding, Blood, Gun, Weapon, Wounded2026.06 | 56.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.68 | — | — | — | — | |
| DiT-PruningResolution=512×512, Sparsity=40%, Params (M)=375.432026.07 | 56.74 | 0.253 | — | — | 0.803 | — | — | — | — | 0.514 | — | — | — | — | — | — | 35.98 | 264.47 | 0.62 | |
| TraSCEBackbone=SD v1.4, Sampler=50-step DDPM, Negative Prompt=Bleeding, Blood, Gun, Weapon, Wounded2026.06 | 56.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.49 | — | — | — | — | |
| MagnitudeResolution=512×512, Sparsity=40%, Params (M)=375.352026.07 | 61.12 | 0.252 | — | — | 0.587 | — | — | — | — | 0.378 | — | — | — | — | — | — | 34.87 | 257.66 | 0.53 | |
| WandaResolution=512×512, Sparsity=20%, Params (M)=494.512026.07 | 61.22 | 0.255 | — | — | 0.879 | — | — | — | — | 0.674 | — | — | — | — | — | — | 36.92 | 269.71 | 0.6 | |
| MagnitudeResolution=512×512, Sparsity=20%, Params (M)=494.292026.07 | 61.25 | 0.257 | — | — | 0.968 | — | — | — | — | 0.484 | — | — | — | — | — | — | 37.76 | 268.88 | 0.59 | |
| WandaResolution=512×512, Sparsity=40%, Params (M)=375.752026.07 | 61.26 | 0.247 | — | — | 0.36 | — | — | — | — | 0.475 | — | — | — | — | — | — | 33.33 | 256.87 | 0.57 | |
| DiT-PruningResolution=512×512, Sparsity=20%, Params (M)=494.342026.07 | 61.71 | 0.257 | — | — | 0.942 | — | — | — | — | 0.693 | — | — | — | — | — | — | 35.78 | 272.1 | 0.6 | |
| DiT-PruningResolution=512×512, Sparsity=50%, Params (M)=315.972026.07 | 61.78 | 0.244 | — | — | 0.285 | — | — | — | — | 0.439 | — | — | — | — | — | — | 33.27 | 254.82 | 0.51 | |
| DenseResolution=512×512, Sparsity=0%, Params (M)=612.082026.07 | 62.36 | 0.256 | — | — | 0.929 | — | — | — | — | — | — | — | — | — | — | — | 36.12 | 274.47 | 0.59 | |
| RECEConflict-Aware=false2026.02 | 65.94 | 0.2811 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RECEConflict-Aware=false, Model modification required=true2026.02 | 67.35 | 0.2767 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT-PruningResolution=256×256, Sparsity=40%, Params (M)=374.552026.07 | 67.5 | 0.26 | — | — | 0.635 | — | — | — | — | 0.484 | — | — | — | — | — | — | 31.06 | 244.21 | 0.51 | |
| Q-DiffusionBackbone=SDXL-UNet, W/A=8/8(int)2025.07 | 68.06 | — | — | — | -1.574 | 0.396 | 16.48 | 0.537 | 13.68 | 0.517 | — | — | — | — | — | — | — | — | — | |
| WandaResolution=256×256, Sparsity=20%, Params (M)=493.642026.07 | 68.6 | 0.264 | — | — | 0.727 | — | — | — | — | 0.631 | — | — | — | — | — | — | 30.62 | 248.03 | 0.5 | |
| DiT-PruningResolution=256×256, Sparsity=20%, Params (M)=493.462026.07 | 69.01 | 0.265 | — | — | 0.793 | — | — | — | — | 0.688 | — | — | — | — | — | — | 31.84 | 249.52 | 0.51 | |
| WandaResolution=256×256, Sparsity=40%, Params (M)=374.872026.07 | 69.32 | 0.256 | — | — | 0.244 | — | — | — | — | 0.456 | — | — | — | — | — | — | 28.86 | 235.79 | 0.47 | |
| MagnitudeResolution=256×256, Sparsity=20%, Params (M)=493.422026.07 | 69.44 | 0.266 | — | — | 0.821 | — | — | — | — | 0.543 | — | — | — | — | — | — | 32.92 | 249.16 | 0.49 | |
| DiT-PruningResolution=256×256, Sparsity=50%, Params (M)=315.102026.07 | 69.79 | 0.251 | — | — | 0.183 | — | — | — | — | 0.385 | — | — | — | — | — | — | 29.76 | 237.6 | 0.42 | |
| MagnitudeResolution=256×256, Sparsity=40%, Params (M)=374.482026.07 | 69.87 | 0.258 | — | — | 0.498 | — | — | — | — | 0.407 | — | — | — | — | — | — | 31.36 | 239.25 | 0.44 | |
| PTQ4DiTBackbone=SD3.5-DiT, W/A=4/8(int)2025.07 | 69.89 | — | — | — | 0.085 | 0.411 | 16.51 | 0.57 | 10.83 | 0.504 | — | — | — | — | — | — | — | — | — | |
| DenseResolution=256×256, Sparsity=0%, Params (M)=612.082026.07 | 70.31 | 0.265 | — | — | 0.807 | — | — | — | — | — | — | — | — | — | — | — | 33.33 | 249.59 | 0.49 | |
| TMPDiffPrecision/Quantization Scheme=Ours, Model=Sana, Resolution=1024x1024, Guidance scale=4.5, Number of samples=10242026.03 | 71.243 | 27.021 | — | — | 1.0434 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| bfloat16Precision/Quantization Scheme=bfloat16, Model=Sana, Resolution=1024x1024, Guidance scale=4.5, Number of samples=10242026.03 | 71.548 | 26.925 | — | — | 0.9476 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| w4a4r32Precision/Quantization Scheme=w4a4r32, Model=Sana, Resolution=1024x1024, Guidance scale=4.5, Number of samples=10242026.03 | 71.839 | 26.907 | — | — | 1.0413 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UCEConflict-Aware=false, Model modification required=true2026.02 | 77.41 | 0.2912 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UCEConflict-Aware=false2026.02 | 78.72 | 0.2909 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WandaResolution=256×256, Sparsity=50%, Params (M)=315.102026.07 | 89.66 | 0.229 | — | — | -0.923 | — | — | — | — | 0.381 | — | — | — | — | — | — | 20.47 | 234.46 | 0.27 |