paper-with-me

홈 › Papers

THINKSAFE: Self-Generated Safety Alignment for Reasoning Models

2026-01-30 · Seanie Lee, Sangwoo Park, Yumin Choi, Gyeongman Kim, Minki Kang, Jihun Yun, Dongmin Park, Jongho Park, Sung Ju Hwang arxiv

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.

📄 PDF Abstract BibTeX arXiv:2601.23143

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation

2026-05-14 · Yu Fu, Longxuan Yu, Haz Sameen Shahgir, Zhipeng Wei 외 arxiv

Safety alignment often improves robustness to harmful queries at the cost of reasoning ability, a tradeoff known as the safety tax. A common cause is distributional mismatch: supervised fine-tuning trains the target mode…

STAIR: Improving Safety Alignment with Introspective Reasoning

2025-02-04 · Yichi Zhang, Siyuan Zhang, Yao Huang, Zeyu Xia 외

Ensuring the safety and harmlessness of Large Language Models (LLMs) has become equally critical as their performance in applications. However, existing safety alignment methods typically suffer from safety-performance t…

Safety Alignment

Stealth Fine-Tuning: Efficiently Breaking Alignment in RVLMs Using Self-Generated CoT

2025-11-18 · Le Yu, Zhengyue Zhao, Yawen Zheng, Yunhao Liu arxiv

Reasoning-augmented Vision-Language Models (RVLMs) rely on safety alignment to prevent harmful behavior, yet their exposed chain-of-thought (CoT) traces introduce new attack surfaces. In this work, we find that the safet…

Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines

2025-11-26 · Yuhang Wang, Yanxu Zhu, Dongyuan Lu, Jitao Sang arxiv

Reasoning models have demonstrated remarkable capabilities in complex reasoning tasks. However, ensuring their safety against adversarial jailbreak prompts remains a critical challenge. Due to the covert and deceptive na…

Large Reasoning Models Learn Better Alignment from Flawed Thinking

2025-10-01 · ShengYun Peng, Eric Smith, Ivan Evtimov, Song Jiang 외 arxiv

Large reasoning models (LRMs) "think" by generating structured chain-of-thought (CoT) before producing a final answer, yet they still lack the ability to reason critically about safety alignment and are easily biased whe…

Reinforcement Learning