Optimizing Decoding Paths in Masked Diffusion Models by Quantifying Uncertainty

2025/12/24 by Chen, Ziyu, Jiang, Xinbei, Sun, Peng +1
#Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2512.21336

Abstract

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.

Citations

• Dream 7B: Diffusion Large Language Models
• DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
• Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking
• Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
• LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models
• Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models
• Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models
• Remasking Discrete Diffusion Models with Inference-Time Scaling
• Large Language Diffusion Models
• Theoretical Benefit and Limitation of Diffusion Language Model
• Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions
• Path Planning for Masked Diffusion Model Sampling
• A General Framework for Inference-time Scaling and Steering of Diffusion Models
• Transformers Can Navigate Mazes With Multi-Step Prediction
• Scaling Diffusion Language Models via Adaptation from Autoregressive Models
• Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning
• Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling
• Discrete Flow Matching
• Simple and Effective Masked Diffusion Language Models
• Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data
• Simplified and Generalized Masked Diffusion for Discrete Data
• The Entropy Enigma: Success and Failure of Entropy Minimization
• Improving Open-Ended Text Generation via Adaptive Decoding
• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution
• Let's Verify Step by Step
• Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation
• The Stable Entropy Hypothesis and Entropy-Aware Decoding: An Analysis and Algorithm for Robust Natural Language Generation
• A Reparameterized Discrete Diffusion Model for Text Generation
• DiffusionBERT: Improving Generative Masked Language Models with Diffusion Models
• SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Control
• MaskGIT: Masked Generative Image Transformer
• High-Resolution Image Synthesis with Latent Diffusion Models
• Training Verifiers to Solve Math Word Problems
• Program Synthesis with Large Language Models
• Structured Denoising Diffusion Models in Discrete State-Spaces
• Evaluating Large Language Models Trained on Code
• Diffusion Models Beat GANs on Image Synthesis
• Understanding Neural Abstractive Summarization Models via Uncertainty
• Denoising Diffusion Implicit Models
• Denoising Diffusion Probabilistic Models
• The sample size required in importance sampling
• Sequence to Sequence Learning with Neural Networks
• Sequence Transduction with Recurrent Neural Networks

Related

• EVODiff: Entropy-aware Variance Optimized Diffusion Inference
• PC-Sampler: Position-Aware Calibration of Decoding Bias in Masked Diffusion Models
• Path Planning for Masked Diffusion Model Sampling
• Unleashing Uncertainty: Efficient Machine Unlearning for Generative AI
• Generative Uncertainty in Diffusion Models