Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models
cs.CL, cs.AI
Submitted: 2026-08-31
Updated: 2026-08-31
Comments: Accepted by EMNLP 2026
Code: https://github.com/Glresearch1/RAEC
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Jailbreaking Black Box Large Language Models in Twenty Queries
- Evaluating Large Language Models Trained on Code
- Training Verifiers to Solve Math Word Problems
- The Llama 3 Herd of Models
- A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models
- Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal
- Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning
- The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
- Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability
- Dream 7B: Diffusion Large Language Models
- Token-level Direct Preference Optimization
- Jailbreaking Large Language Diffusion Models: Revealing Hidden Safety Flaws in Diffusion-Based Text Generation
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering