# LogicDiff: Logic-Guided Denoising Improves Zero-Shot Reasoning in Masked Diffusion Language Models

arXiv: https://arxiv.org/abs/2603.26771 · HTML: https://mlsanity.com/paper/2603.26771/

## Key facts

- Cited by 1 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 9 of its 17 references that are in the map.
- First posted on arXiv 24 Mar 2026.

## Placed in

- LLM reasoning / [Diffusion reasoning and test-time scaling](https://mlsanity.com/lr/diffusion-reasoning-and-test-time-scaling/)
- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)

## Builds on (most cited first, 9 of 9)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning](https://mlsanity.com/paper/2504.12216/) (2025)
- [The Flexibility Trap: Rethinking the Value of Arbitrary Order in Diffusion Language Models](https://mlsanity.com/paper/2601.15165/) (2026)
- [Advancing Reasoning in Diffusion Language Models with Denoising Process Rewards](https://mlsanity.com/paper/2510.01544/) (2025)
- [Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation](https://mlsanity.com/paper/2601.13228/) (2026)
- [Reasoning in Diffusion Large Language Models is Concentrated in Dynamic Confusion Zones](https://mlsanity.com/paper/2511.15208/) (2025)
- [Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching](https://mlsanity.com/paper/2602.22871/) (2026)
- [DOS: Dependency-Oriented Sampler for Masked Diffusion Language Models](https://mlsanity.com/paper/2603.15340/) (2026)

## Built on by (most cited first, 1 of 1)

- [The Confidence Shortcut: A Reasoning Failure Mode of Masked Diffusion Models](https://mlsanity.com/paper/2605.29123/) (2026)

## Abstract (arXiv)

Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens from a fully masked sequence. Their standard confidence-based unmasking strategy systematically defers high-entropy logical connective tokens, degrading reasoning performance. We introduce LogicDiff, an inference-time method that replaces confidence-based unmasking with logic-role-guided unmasking. A lightweight classification head (4.2M parameters, 0.05% of the base model) predicts the logical role of each masked position (premise, connective, derived step, conclusion, or filler) from the base model's hidden states with 98.4% accuracy, and a dependency-ordered scheduler unmasks tokens in logical order. In zero-shot settings, LogicDiff improves LLaDA-8B-Instruct accuracy from 22.0% to 60.7% on GSM8K (+38.7 percentage points) and from 23.6% to 29.2% on MATH-500 (+5.6 pp), with less than 6% speed overhead. However, with 8-shot chain-of-thought prompting, the baseline reaches approximately 70% and LogicDiff provides no additional improvement. Analysis reveals that few-shot prompting implicitly resolves the same ordering problem that LogicDiff explicitly addresses, and that fixed role-based ordering can cause premature commitment to numerical values before sufficient context is available. Our results characterize the Flexibility Trap as primarily a zero-shot phenomenon and identify context-adaptive ordering as a key direction for future work.