DARE: Diffusion Large Language Models Alignment and Reinforcement Executor
arXiv:2604.04215v1 Announce Type: new Abstract: Diffusion large language models (dLLMs) are emerging as a compelling alternative to dominant autoregressive models, replacing strictly sequential token generation …
Jingyi Yang, Yuxian Jiang, Xuhao Hu, Shuang Cheng, Biqing Qi, Jing Shao
20 views