Revise, Don't Freeze: Sampler-Matched Training for Self-Correcting Masked Diffusion Language Models 事件

Revise, Don't Freeze: Sampler-Matched Training for Self-Correcting Masked Diffusion Language Models arXiv:2606.01026v1 Announce Type: new Abstract: Masked diffusion language models (MDLMs) re-predict every position at each denoising step, but standard samplers commit tokens once revealed, leaving this revision capability unused. Existing approaches either add heuristic or learned mechanisms to revise committed tokens, or remask them back to [MASK] before re-predicting; a principled sampler that