Exploring Discrete Diffusion Models for Image Captioning