titleNemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation
authorsShufan Li, Hanrong Ye, Greg Heinrich, Yonggan Fu, Haotian Zhang, Sifei Liu, Aditya Grover, Pavlo Molchanov
date2026-10-27
venuearXiv (preprint, not peer reviewed)
idarXiv:2610.36466
affiliationNVIDIA
upvotes30 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
abstractWe propose Nemotron-Labs-Diffusion-Uni (NLD-Uni), a unified discrete diffusion large language model (DLLM) for visual understanding and generation tasks. Compared with prior unified diffusion models, NLD-Uni introduces two major innovations. First, we propose a unified cross-paradigm on-policy distillation (OPD) framework that successfully transfers the visual understanding ability of autoregressive VLMs and visual generation ability of continuous flow models, to a unified DLLM. Second, NLD-Uni incorporates a novel progressive self-refining diffusion process that gradually corrupts each token across multiple steps during training, as opposed to adopting the conventional masked diffusion formulation that directly corrupts a clean image token to an uninformative mask token. At inference, this design allows NLD-Uni to make intermediate commit to a token, enabling progressive self-refinement for better visual quality. Through extensive experiments, we show that NLD-Uni achieves state-of-the-art performance among unified DLLMs on diverse tasks including visual question-answering, multi-modal reasoning, object grounding, text-to-image generation, and image editing.

cite

@misc{li2026nemotron,
  title={Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation},
  author={Shufan Li and Hanrong Ye and Greg Heinrich and Yonggan Fu and Haotian Zhang and Sifei Liu and Aditya Grover and Pavlo Molchanov},
  year={2026},
  eprint={2610.36466},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2610.36466}
}