Abstract: Recent diffusion-based models have demonstrated marked advancement in synthesizing high-quality images from textual prompts. However, they often struggle to follow user-specified spatial ...