Beyond Hard Masks: Progressive Token Evolution for Diffusion Language Models

TL;DR

EvoToken-DLM introduces a diffusion-based language modeling method that replaces hard masks with evolving soft token distributions, enabling revisable decoding and achieving superior performance over existing diffusion and masked models.

Contribution

The paper presents EvoToken-DLM, a novel approach that replaces binary masks with soft token distributions and introduces continuous trajectory supervision for improved language modeling.

Findings

Outperforms existing diffusion-based models

Supports revisable decoding with soft token evolution

Achieves superior results across multiple benchmarks

Abstract

Diffusion Language Models (DLMs) offer a promising alternative for language modeling by enabling parallel decoding through iterative refinement. However, most DLMs rely on hard binary masking and discrete token assignments, which hinder the revision of early decisions and underutilize intermediate probabilistic representations. In this paper, we propose EvoToken-DLM, a novel diffusion-based language modeling approach that replaces hard binary masks with evolving soft token distributions. EvoToken-DLM enables a progressive transition from masked states to discrete outputs, supporting revisable decoding. To effectively support this evolution, we introduce continuous trajectory supervision, which aligns training objectives with iterative probabilistic updates. Extensive experiments across multiple benchmarks show that EvoToken-DLM consistently achieves superior performance, outperforming strong diffusion-based and masked DLM baselines. Project webpage: https://aim-uofa.github.io/EvoTokenDLM.