# Rethinking Reasoning with MDLMs: Early Exits, Post-hoc Reasoning, and Beyond

arXiv: https://arxiv.org/abs/2510.19990 · HTML: https://mlsanity.com/paper/2510.19990/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion reasoning and test-time scaling, one of 17 directions in LLM reasoning.
- Builds on 33 of its 53 references that are in the map.
- First posted on arXiv 22 Oct 2025.

## Placed in

- LLM reasoning / [Diffusion reasoning and test-time scaling](https://mlsanity.com/lr/diffusion-reasoning-and-test-time-scaling/)
- Efficient LLM inference and serving / [Efficient reasoning and agent inference](https://mlsanity.com/inference/efficient-reasoning-and-agent-inference/)

## Builds on (most cited first, 30 of 33)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [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)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [Simplified and Generalized Masked Diffusion for Discrete Data](https://mlsanity.com/paper/2406.04329/) (2024)
- [Dream 7B: Diffusion Large Language Models](https://mlsanity.com/paper/2508.15487/) (2025)
- [Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding](https://mlsanity.com/paper/2505.22618/) (2025)
- [The Lessons of Developing Process Reward Models in Mathematical Reasoning](https://mlsanity.com/paper/2501.07301/) (2025)
- [Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models](https://mlsanity.com/paper/2503.09573/) (2025)
- [Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data](https://mlsanity.com/paper/2406.03736/) (2024)
- [Scaling Diffusion Language Models via Adaptation from Autoregressive Models](https://mlsanity.com/paper/2410.17891/) (2024)
- [A General Framework for Inference-time Scaling and Steering of Diffusion Models](https://mlsanity.com/paper/2501.06848/) (2025)
- [Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions](https://mlsanity.com/paper/2502.06768/) (2025)
- [DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation](https://mlsanity.com/paper/2506.20639/) (2025)
- [d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning](https://mlsanity.com/paper/2504.12216/) (2025)
- [The pitfalls of next-token prediction](https://mlsanity.com/paper/2403.06963/) (2024)
- [Simple Guidance Mechanisms for Discrete Diffusion Models](https://mlsanity.com/paper/2412.10193/) (2024)
- [Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking](https://mlsanity.com/paper/2505.24857/) (2025)
- [OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models](https://mlsanity.com/paper/2411.04905/) (2024)
- [Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs](https://mlsanity.com/paper/2402.14740/) (2024)
- [Reinforcing General Reasoning without Verifiers](https://mlsanity.com/paper/2505.21493/) (2025)
- [Diffusion Language Models Know the Answer Before Decoding](https://mlsanity.com/paper/2508.19982/) (2025)
- [Diffusion Beats Autoregressive in Data-Constrained Settings](https://mlsanity.com/paper/2507.15857/) (2025)
- [Edit Flows: Flow Matching with Edit Operations](https://mlsanity.com/paper/2506.09018/) (2025)
- [Any-Order Flexible Length Masked Diffusion](https://mlsanity.com/paper/2509.01025/) (2025)

## Built on by (most cited first, 4 of 4)

- [Follow the Latent Roadmap: Navigating Revocable Decoding for Diffusion LLMs with Anchor Tokens](https://mlsanity.com/paper/2606.16847/) (2026)
- [Reasoning or Rationalization? The Role of Justifications in Masked Diffusion Models for Fact Verification](https://mlsanity.com/paper/2603.01190/) (2026)
- [Detect, Remask, Repair: Diffusion Editing for Faithful Summarization of Evolving Contexts](https://mlsanity.com/paper/2606.12807/) (2026)
- [Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models](https://mlsanity.com/paper/2610.03665/) (2026)

## Abstract (arXiv)

The reasoning paradigm, where language models reason before answering, has enabled breakthroughs on tasks such as mathematical problem-solving. While current tooling for reasoning is built around next-token prediction trained models, recent works introduce an alternative choice: masked diffusion language models (MDLMs). MDLMs are trained to infill positions in randomly masked sequences. We introduce reasoning-as-infilling, a prompting technique that pre-fills tokens to explicitly delimit reasoning and answer regions, unlocking a unified set of capabilities for MDLM reasoning. Because answer positions are explicitly designated, the model's conditional distributions over answer tokens are directly accessible during generation. This enables early exits when the model is certain of its answer. The same framework supports post-hoc reasoning: given question-answer pairs, MDLMs can sample high-quality reasoning traces from their posterior, a distribution that is intractable for autoregressive models. On GSM8k, fine-tuning LLaDA-8B-Base on these posterior traces improves accuracy by +14.9%, matching gains from human-written traces (+13.4%). Finally, given a reference answer, the answer region distributions enable scoring partial reasoning traces at intermediate steps, providing intermediate rewards that are more strongly correlated with correctness than scores from a specialized process reward model. At intermediate steps, answer-likelihood scores from autoregressive models are significantly less predictive of correctness than those from MDLMs. Our results demonstrate that the MDLM training objective provides promising benefits for reasoning.