arXivLongxuan Yu, Yunshu Wu, Yu Fu, Siheng Xiong, Rob Brekelmans, Hui Liu, Yue Dong, Greg Ver SteegSat, May 30, 2026, 10:27 PM PDT
score 16.0
Masked language model learns to generate text smoothly without repeated words
Original: DSL-LLaDA: Scaling Continuous Denoising to 8B Masked Diffusion LMs
Source: arxiv.org ↗
Writing ELI5 summary…