Thinking tokens in reasoning models do not enable safety deliberation; refusal/compliance is strongly predictable from the first token and rarely changes during thinking.
THINKSAFE: Self-Generated Safety Alignment for Reasoning Models
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
abstract
Large reasoning models (LRMs) achieve remarkable performance by leveraging reinforcement learning (RL) on reasoning tasks to generate long chain-of-thought (CoT) reasoning. However, this over-optimization often prioritizes compliance, making models vulnerable to harmful prompts. To mitigate this safety degradation, recent approaches rely on external teacher distillation, yet this introduces a distributional discrepancy that degrades native reasoning. We formalize safety realignment as a KL projection onto the safe simplex and prove that the student's own safety-filtered distribution is the unique KL-optimal target, while any external teacher incurs an irreducible excess KL penalty. Guided by this analysis, we propose ThinkSafe, a self-generated alignment framework that restores safety without external teachers. Our key insight is that while compliance suppresses safety mechanisms, models often retain latent knowledge to identify harm. ThinkSafe unlocks this via lightweight refusal steering, which preserves the KL-optimal target while increasing the acceptance rate. Experiments on DeepSeek-R1-Distill and Qwen3 show ThinkSafe significantly improves safety while preserving reasoning proficiency, and achieves superior safety and comparable reasoning to GRPO with roughly an order of magnitude less compute. Code, models, and datasets are available at https://github.com/seanie12/ThinkSafe and https://huggingface.co/Seanie-lee/collections.
years
2026 4representative citing papers
MCPO applies contrastive learning to GRPO-style RL by treating cross-domain correct rollouts as positives and incorrect ones as negatives to improve multi-domain reasoning performance in LRMs.
On-policy self-distillation with teacher flip rate yields better safety-reasoning tradeoffs than off-policy or external-teacher baselines across model scales.
SELFCI uses complementary self-distillation with two reverse KL divergences to align LLMs to contextual integrity while preserving utility, outperforming RL baselines like GRPO in agentic settings.
citing papers explorer
-
Do Thinking Tokens Help with Safety?
Thinking tokens in reasoning models do not enable safety deliberation; refusal/compliance is strongly predictable from the first token and rarely changes during thinking.
-
Harmony in Diversity: Multi-domain Contrastive Policy Optimization for Large Reasoning Models
MCPO applies contrastive learning to GRPO-style RL by treating cross-domain correct rollouts as positives and incorrect ones as negatives to improve multi-domain reasoning performance in LRMs.
-
Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation
On-policy self-distillation with teacher flip rate yields better safety-reasoning tradeoffs than off-policy or external-teacher baselines across model scales.
-
It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs
SELFCI uses complementary self-distillation with two reverse KL divergences to align LLMs to contextual integrity while preserving utility, outperforming RL baselines like GRPO in agentic settings.