Blockwise SFT for Diffusion Language Models: Reconciling Bidirectional Attention and Autoregressive Decoding
机构 * University of California, Merced(加州大学默塞德分校) ; The University of Queensland(昆士兰大学) ; Google DeepMind(谷歌DeepMind)
专题命中 指令微调 :language model(title,abstract);SFT(title,abstract);分类 cs.CL