# Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models

arXiv: https://arxiv.org/abs/2609.00495 · HTML: https://mlsanity.com/paper/2609.00495/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 26 of its 35 references that are in the map.
- First posted on arXiv 31 Aug 2026.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)
- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- LLM safety and security / [Jailbreaking](https://mlsanity.com/safety/jailbreaking/)

## Builds on (most cited first, 26 of 26)

- [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)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Structured Denoising Diffusion Models in Discrete State-Spaces](https://mlsanity.com/paper/2107.03006/) (2021)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Jailbreaking Black Box Large Language Models in Twenty Queries](https://mlsanity.com/paper/2310.08419/) (2023)
- [Diffusion-LM Improves Controllable Text Generation](https://mlsanity.com/paper/2205.14217/) (2022)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [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)
- [Dream 7B: Diffusion Large Language Models](https://mlsanity.com/paper/2508.15487/) (2025)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [A Continuous Time Framework for Discrete Denoising Models](https://mlsanity.com/paper/2205.14987/) (2022)
- [A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily](https://mlsanity.com/paper/2311.08268/) (2023)
- [DiffusionBERT: Improving Generative Masked Language Models with Diffusion Models](https://mlsanity.com/paper/2211.15029/) (2022)
- [SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Control](https://mlsanity.com/paper/2210.17432/) (2022)
- [SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities](https://mlsanity.com/paper/2502.12025/) (2025)
- [STAR-1: Safer Alignment of Reasoning LLMs with 1K Data](https://mlsanity.com/paper/2504.01903/) (2025)
- [The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs](https://mlsanity.com/paper/2507.11097/) (2025)
- [Jailbreaking Large Language Diffusion Models: Revealing Hidden Safety Flaws in Diffusion-Based Text Generation](https://mlsanity.com/paper/2507.19227/) (2025)
- [Nudging: Inference-time Alignment of LLMs via Guided Decoding](https://mlsanity.com/paper/2410.09300/) (2024)
- [Where to Start Alignment? Diffusion Large Language Model May Demand a Distinct Position](https://mlsanity.com/paper/2508.12398/) (2025)
- [Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability](https://mlsanity.com/paper/2510.00565/) (2025)
- [A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models](https://mlsanity.com/paper/2509.23286/) (2025)
- [Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning](https://mlsanity.com/paper/2603.07445/) (2026)

## Abstract (arXiv)

Diffusion large language models (dLLMs) generate text through iterative denoising rather than left-to-right decoding. This generation paradigm introduces two axes that can influence safety alignment: when tokens are generated during denoising and where they appear in the response. In this paper, we measure dLLM safety behavior under harmful prompts by tracing intermediate token distributions and commitment decisions throughout denoising. Our analysis shows that refusal signals are concentrated in early denoising steps and leading response positions, and the tokens committed early can strongly shape the final safety outcome. Our measurements further show that the denoising step and persistence of refusal-token commitment are important for understanding dLLM safety. Based on these findings, we propose Refusal-Aware Early Commitment (RAEC), a simple training-free decoding method that commits persistent refusal signals from early steps. Experiments on LLaDA and Dream show that RAEC reduces attack success rates while largely preserving utility. The code is available at https://github.com/Glresearch1/RAEC.