Unlocking Prompt Infilling Capability for Diffusion Language Models

Yoshinari Fujinuma, Keisuke Sakaguchi · 2026

Read Unlocking Prompt Infilling Capability for Diffusion Language Models by Yoshinari Fujinuma, Keisuke Sakaguchi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-03677.txt.

Description

Yoshinari Fujinuma and Keisuke Sakaguchi unlocked a missing skill in masked diffusion language models, which build text by cleaning up masked passages in both directions. Standard tuning masks only answers, so the models could not fill gaps inside a prompt; masking prompts and answers together during tuning fixed that. Model-filled prompts matched or beat hand-written ones and carried across models.

Catalog record

CREATORYoshinari Fujinuma, Keisuke Sakaguchi
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43909
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.03677

Browse the full library · About the Tower of Babel