# Improving Discrete Diffusion Unmasking Policies Beyond Explicit Reference Policies

arXiv: https://arxiv.org/abs/2510.05725 · HTML: https://mlsanity.com/paper/2510.05725/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 30 of its 37 references that are in the map.
- First posted on arXiv 7 Oct 2025.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)

## Builds on (most cited first, 30 of 30)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Elucidating the Design Space of Diffusion-Based Generative Models](https://mlsanity.com/paper/2206.00364/) (2022)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [Structured Denoising Diffusion Models in Discrete State-Spaces](https://mlsanity.com/paper/2107.03006/) (2021)
- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [Simple and Effective Masked Diffusion Language Models](https://mlsanity.com/paper/2406.07524/) (2024)
- [Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution](https://mlsanity.com/paper/2310.16834/) (2023)
- [Simplified and Generalized Masked Diffusion for Discrete Data](https://mlsanity.com/paper/2406.04329/) (2024)
- [Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions](https://mlsanity.com/paper/2102.05379/) (2021)
- [Dream 7B: Diffusion Large Language Models](https://mlsanity.com/paper/2508.15487/) (2025)
- [A Continuous Time Framework for Discrete Denoising Models](https://mlsanity.com/paper/2205.14987/) (2022)
- [Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data](https://mlsanity.com/paper/2406.03736/) (2024)
- [LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models](https://mlsanity.com/paper/2505.19223/) (2025)
- [Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions](https://mlsanity.com/paper/2502.06768/) (2025)
- [Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling](https://mlsanity.com/paper/2409.02908/) (2024)
- [d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning](https://mlsanity.com/paper/2504.12216/) (2025)
- [Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference](https://mlsanity.com/paper/2508.02193/) (2025)
- [Reinforcement Learning with Verifiable Rewards: GRPO's Effective Loss, Dynamics, and Success Amplification](https://mlsanity.com/paper/2503.06639/) (2025)
- [Training and Inference on Any-Order Autoregressive Models the Right Way](https://mlsanity.com/paper/2205.13554/) (2022)
- [Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models](https://mlsanity.com/paper/2505.10446/) (2025)
- [Causal Language Modeling Can Elicit Search and Reasoning Capabilities on Logic Puzzles](https://mlsanity.com/paper/2409.10502/) (2024)
- [Fine-Tuning Discrete Diffusion Models with Policy Gradient Methods](https://mlsanity.com/paper/2502.01384/) (2025)
- [PPO-Clip Attains Global Optimality: Towards Deeper Understandings of Clipping](https://mlsanity.com/paper/2312.12065/) (2023)
- [Discovering Non-monotonic Autoregressive Orderings with Variational Inference](https://mlsanity.com/paper/2110.15797/) (2021)

## Built on by (most cited first, 22 of 22)

- [Learning Unmasking Policies for Diffusion Language Models](https://mlsanity.com/paper/2512.09106/) (2025)
- [Stop Training for the Worst: Progressive Unmasking Accelerates Masked Diffusion Training](https://mlsanity.com/paper/2602.10314/) (2026)
- [Where-to-Unmask: Ground-Truth-Guided Unmasking Order Learning for Masked Diffusion Language Models](https://mlsanity.com/paper/2602.09501/) (2026)
- [Unifying Masked Diffusion Models with Various Generation Orders and Beyond](https://mlsanity.com/paper/2602.02112/) (2026)
- [Locally Confident, Globally Stuck: The Quality-Exploration Dilemma in Diffusion Language Models](https://mlsanity.com/paper/2604.00375/) (2026)
- [Fixed-Point Masked Generative Modeling](https://mlsanity.com/paper/2605.31215/) (2026)
- [Reinforcement Learning for Diffusion LLMs with Entropy-Guided Step Selection and Stepwise Advantages](https://mlsanity.com/paper/2603.12554/) (2026)
- [Attention-Discounted Adaptive Sampler for Masked Diffusion Language Models](https://mlsanity.com/paper/2606.10829/) (2026)
- [Tuning the Implicit Regularizer of Masked Diffusion Language Models: Enhancing Generalization via Insights from k-Parity](https://mlsanity.com/paper/2601.22450/) (2026)
- [The Confidence Shortcut: A Reasoning Failure Mode of Masked Diffusion Models](https://mlsanity.com/paper/2605.29123/) (2026)
- [Scheduling Thoughts: Learning the Order of Thought in Diffusion Language Models](https://mlsanity.com/paper/2606.23567/) (2026)
- [From Interface to Inference: Eliciting Any-Order Inference from Any-Order Models](https://mlsanity.com/paper/2607.26504/) (2026)
- [Learning Generation Orders for Masked Discrete Diffusion Models via Variational Inference](https://mlsanity.com/paper/2602.23968/) (2026)
- [Progressive Refinement Regulation for Accelerating Diffusion Language Model Decoding](https://mlsanity.com/paper/2603.04514/) (2026)
- [Understanding and Accelerating the Training of Masked Diffusion Language Models](https://mlsanity.com/paper/2605.13026/) (2026)
- [Adaptive Order Policies for Masked Diffusion](https://mlsanity.com/paper/2606.00295/) (2026)
- [One-step lowest-variance selection in a Gaussian random-field model motivated by masked diffusion: Total correlation and a square root collision threshold](https://mlsanity.com/paper/2607.17522/) (2026)
- [DDSynth-RL: Audio Synthesizer Inversion via Discrete Diffusion with Reinforcement Learning](https://mlsanity.com/paper/2608.03032/) (2026)
- [Entropy Can Flow, or It Can Guide. Be Entropy. LEDFlow: Introducing Entropy-guided Generation Order into Uniform Discrete Flow](https://mlsanity.com/paper/2609.25131/) (2026)
- [Optimizing Denoising Trajectories in dLLMs: A Lightweight Evolutionary Heuristic Approach](https://mlsanity.com/paper/2609.26052/) (2026)
- [Does Uniform Discrete Diffusion Need Time?](https://mlsanity.com/paper/2609.30977/) (2026)
- [Batch-Paced: Fixed-Budget Batch Decoding in Discrete Diffusion Language Models](https://mlsanity.com/paper/2610.17917/) (2026)

## Abstract (arXiv)

Masked diffusion models (MDMs) have recently emerged as a novel framework for language modeling. MDMs generate sentences by iteratively denoising masked sequences, filling in [MASK] tokens step by step. Although MDMs support any-order sampling, performance is highly sensitive to the choice of which position to unmask next. Prior work typically relies on rule-based schedules (e.g., max-confidence, max-margin), which provide ad hoc improvements. In contrast, we replace these heuristics with a learned scheduler. Specifically, we cast denoising as a KL-regularized Markov decision process (MDP) with an explicit reference policy and optimize a regularized objective that admits policy improvement and convergence guarantees under standard assumptions. We prove that the optimized policy under this framework generates samples that more closely match the data distribution than heuristic schedules. Empirically, across four benchmarks, our learned policy consistently outperforms max-confidence: for example, on SUDOKU, where unmasking order is critical, it yields a 20.1% gain over random and a 11.2% gain over max-confidence. Code is available at https://github.com/chunsanHong/UPO.