Unified multimodal diffusion large language models (dLLMs) offer a single architecture for both image generation and multimodal understanding, but their iter…
机构:Meta
来源:arXiv 2610.10990 | AI4Papers 论文推荐平台