Saved in:
Bibliographic Details
Main Authors: Wu, Yue, Huang, Jian
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.17068
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914486653288448
author Wu, Yue
Huang, Jian
author_facet Wu, Yue
Huang, Jian
contents Diffusion large language models (dLLMs) enable parallel text generation by iteratively denoising a fully masked sequence, unmasking a subset of masked tokens at each step. Existing decoding strategies rely on static confidence metrics computed at a single denoising step, ignoring temporal history and often leading to premature unmasking of unstable tokens. In this work, we theoretically establish that a token's temporal instability, quantified by the KL divergence between consecutive prediction distributions, provides a strict lower bound on its mutual information with the remaining masked context, indicating that temporally unstable tokens are inherently unsafe to unmask. Based on this insight, we propose Stability-Weighted Decoding (SWD), a training-free, plug-and-play strategy that incorporates temporal stability into token scoring and acts as a universal modulator for arbitrary score-based decoding policies. Experiments on code generation and mathematical reasoning benchmarks demonstrate that SWD consistently improves generation accuracy across representative scoring metrics and selection policies, and exhibits exceptional robustness, maintaining a significant performance lead over standard baselines across varying acceleration ratios.
format Preprint
id arxiv_https___arxiv_org_abs_2604_17068
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stability-Weighted Decoding for Diffusion Language Models
Wu, Yue
Huang, Jian
Computation and Language
Machine Learning
Diffusion large language models (dLLMs) enable parallel text generation by iteratively denoising a fully masked sequence, unmasking a subset of masked tokens at each step. Existing decoding strategies rely on static confidence metrics computed at a single denoising step, ignoring temporal history and often leading to premature unmasking of unstable tokens. In this work, we theoretically establish that a token's temporal instability, quantified by the KL divergence between consecutive prediction distributions, provides a strict lower bound on its mutual information with the remaining masked context, indicating that temporally unstable tokens are inherently unsafe to unmask. Based on this insight, we propose Stability-Weighted Decoding (SWD), a training-free, plug-and-play strategy that incorporates temporal stability into token scoring and acts as a universal modulator for arbitrary score-based decoding policies. Experiments on code generation and mathematical reasoning benchmarks demonstrate that SWD consistently improves generation accuracy across representative scoring metrics and selection policies, and exhibits exceptional robustness, maintaining a significant performance lead over standard baselines across varying acceleration ratios.
title Stability-Weighted Decoding for Diffusion Language Models
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2604.17068