2512.21336

Total: 1

#1 Optimizing Decoding Paths in Masked Diffusion Models by Quantifying Uncertainty [PDF4] [Copy] [Kimi9] [REL]

Authors: Ziyu Chen, Xinbei Jiang, Peng Sun, Tao Lin

Masked Diffusion Models (MDMs) offer flexible, non-autoregressive generation, but this freedom introduces a challenge: final output quality is highly sensitive to the decoding order. We are the first to formalize this issue, attributing the variability in output quality to the cumulative predictive uncertainty along a generative path. To quantify this uncertainty, we introduce Denoising Entropy, a computable metric that serves as an internal signal for evaluating generative process. Leveraging this metric, we propose two algorithms designed to optimize the decoding path: a post-hoc selection method and a real-time guidance strategy. Experiments demonstrate that our entropy-guided methods significantly improve generation quality, consistently boosting accuracy on challenging reasoning, planning, and code benchmarks. Our work establishes Denoising Entropy as a principled tool for understanding and controlling generation, effectively turning the uncertainty in MDMs from a liability into a key advantage for discovering high-quality solutions.

Subjects: Computation and Language , Artificial Intelligence , Machine Learning

Publish: 2025-12-24 18:59:51 UTC