Lookahead Unmasking Elicits Accurate Decoding in Diffusion Language Models

Sanghyun Lee¹ · Seungryong Kim¹ · Jongho Park² · Dongmin Park^3†

¹ KAIST · ² University of California, Berkeley · ³ KRAFTON

^† Corresponding author.

Paper | Project Page

Code coming soon.