Announce our new work, SILMM (Self-Improving Large Multimodal Models for Compositional Text-to-Image Generation)!