LaViDa: A Large Diffusion Language Model for Multimodal Understanding | Read Paper on Bytez