Beyond Hard Masks: Progressive Token Evolution for Diffusion Language Models
This work addresses a bottleneck in diffusion language models for natural language processing, offering an incremental improvement by enabling revisable decoding and better utilization of probabilistic representations.
The paper tackles the problem of diffusion language models being hindered by hard binary masking and discrete token assignments, which limit revision of early decisions and underutilize intermediate representations, by proposing EvoToken-DLM that uses evolving soft token distributions and continuous trajectory supervision, resulting in superior performance across multiple benchmarks compared to baselines.
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.