Search for a command to run...
From Denoising to Refining: A Corrective Framework for Vision-Language Diffusion Model