yomi-chen's picture
Publish run3: six models,1200 outputs and EditableDesignBench v4 evaluation data
15fe5b5 verified
|
Raw History Blame Contribute Delete
2.13 kB

Run3: generation and pixel metrics

Pixel-only. MLLM judging is not requested or executed. No semantic scores or joint qualification are reported.

Generated: 1200/1200; generation errors: 0; pending: 0. Pixel-valid cases: 1200/1200. Requested image/pixel scope complete: True.

Pixel means appear only after the formal report validates full coverage for that metric. Text and assets are reported separately.

Configuration notice — joyai-image-edit-plus: Official default bucket output,30steps,CFG4. Original source input without padding; omit width/height. Raw output is preserved. Pixel references/masks are mapped to the official output geometry. These pixel scores use a different resolution protocol from the other five models; report separately.

text

Model Images Errors Pending Pixel valid MSE ↓ PSNR ↑ SSIM ↑ LPIPS ↓
fibo2-edit-turbo 100/100 0 0 100/100 63.7251 32.6149 0.9888 0.0097
joyai-image-edit-plus 100/100 0 0 100/100 637.8287 23.2513 0.8378 0.0567
qwen-image21 100/100 0 0 100/100 563.8236 21.9971 0.8156 0.0500
qwen-image21-turbo 100/100 0 0 100/100 525.3037 22.3523 0.8267 0.0435
flux3-klein 100/100 0 0 100/100 239.6809 27.6982 0.9419 0.0269
sensenova-u1.5-8b-mot 100/100 0 0 100/100 236.0134 25.8899 0.9430 0.0625

assets

Model Images Errors Pending Pixel valid MSE ↓ PSNR ↑ SSIM ↑ LPIPS ↓
fibo2-edit-turbo 100/100 0 0 100/100 137.6570 30.4272 0.9825 0.0141
joyai-image-edit-plus 100/100 0 0 100/100 899.8613 21.7411 0.8341 0.0644
qwen-image21 100/100 0 0 100/100 534.1776 22.3092 0.8343 0.0493
qwen-image21-turbo 100/100 0 0 100/100 425.0533 23.3202 0.8452 0.0429
flux3-klein 100/100 0 0 100/100 228.0066 28.0610 0.9504 0.0215
sensenova-u1.5-8b-mot 100/100 0 0 100/100 308.9311 24.9309 0.9301 0.0677