Model Architecture & Training MethodsShen Nie et al. (Renmin University of China, Ant Group)Renmin University of ChinaAnt Group
Nie et al. present LLaDA
Presented LLaDA — an 8B diffusion language model rivals LLaMA3 8B at in-context learning and beats GPT-4o at reversal completion
Why it matters
Reopened the question of whether autoregression is a necessary condition for LLM capabilities at all
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/2502.09992Added in Week 35, 2026 · See that week