Wow, really fast!

Researchers at the National University of Singapore present DMax!

This new paradigm for diffusion language models (dLLMs) enables aggressive parallel decoding.

It cleverly mitigates error accumulation by reforming decoding as a progressive self-refinement… pic.twitter.com/T7ZS5jO4sr

— 机器之心 JIQIZHIXIN (@jiqizhixin) 10 avril 2026

Wow, really fast! Researchers at the National University of Singapore present DMax! This new paradigm for diffusion language models (dLLMs) enables aggressive parallel decoding. It cleverly mitigates error accumulation by reforming decoding as a progressive self-refinement

→ View original post on X — @jiqizhixin