Fetching the paper…

CONFORM: Contrast is All You Need For High-Fidelity Text-to-Image Diffusion Models · Around