Reasoning Structure Matters for Safety Alignment of Reasoning Models

arXiv:2604.18946v1 Announce Type: new Abstract: Large reasoning models (LRMs) achieve strong performance on complex reasoning tasks but often generate harmful responses to malicious user queries. This paper investigates the underlying cause of these safety risks and shows that the issue lies in the reasoning structure itself. Based on this insight, we claim that effective safety alignment can be achieved by altering the reasoning structure. We propose AltTrain, a simple yet effective post training method that explicitly alters the reasoning structure of LRMs. AltTrain is both practical and generalizable, requiring no complex reinforcement learning (RL) training or reward design, only supervised finetuning (SFT) with a lightweight 1K training examples. Experiments across LRM backbones and model sizes demonstrate strong safety alignment, along with robust generalization across reasoning, QA, summarization, and multilingual setting.

Related

• Think Less, Know More: State-Aware Reasoning Compression with Knowledge Guidance for Efficient Reasoning
• Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model
• StaRPO: Stability-Augmented Reinforcement Policy Optimization
• SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks
• ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning

Source: arXiv cs.AI | 2026-04-22