CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models
arXiv:2610.08833v1 Announce Type: new Abstract: Masked diffusion language models (MDLMs) decode by repeatedly committing tokens to masked positions, but these commitments are usually irreversible. A token chosen under sparse, partial context is kept fixed, even when later context no longer supports…
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-10-08 04:00 · arXiv cs.CL
CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models