paper-with-me

Papers

UnsafeChain: Enhancing Reasoning Model Safety via Hard Cases

2025-07-29 · Raj Vardhan Tomar, Preslav Nakov, Yuxia Wang arxiv

As large reasoning models (LRMs) grow more capable, chain-of-thought (CoT) reasoning introduces new safety challenges. Existing SFT-based safety alignment studies dominantly focused on filtering prompts with safe, high-quality responses, while overlooking hard prompts that always elicit harmful outputs. To fill this gap, we introduce UnsafeChain, a safety alignment dataset constructed from hard prompts with diverse sources, where unsafe completions are identified and explicitly corrected into safe responses. By exposing models to unsafe behaviors and guiding their correction, UnsafeChain enhances safety while preserving general reasoning ability. We fine-tune three LRMs on UnsafeChain and compare them against recent SafeChain and STAR-1 across six out-of-distribution and five in-distribution benchmarks. UnsafeChain consistently outperforms prior datasets, with even a 1K subset matching or surpassing baseline performance, demonstrating the effectiveness and generalizability of correction-based supervision. We release our dataset and code at https://github.com/mbzuai-nlp/UnsafeChain

📄 PDF Abstract BibTeX arXiv:2507.21652

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Enhancing RL Safety with Counterfactual LLM Reasoning

2024-09-16 · Dennis Gross, Helge Spieker

Reinforcement learning (RL) policies may exhibit unsafe behavior and are hard to explain. We use counterfactual large language model reasoning to enhance RL policy safety post-training. We show that our approach improves…

counterfactualLanguage ModelingLanguage ModellingLarge Language Model+3

How Should We Enhance the Safety of Large Reasoning Models: An Empirical Study

2025-05-21 · Zhexin Zhang, Xian Qi Loye, Victor Shea-Jay Huang, Junxiao Yang 외

Large Reasoning Models (LRMs) have achieved remarkable success on reasoning-intensive tasks such as mathematics and programming. However, their enhanced reasoning capabilities do not necessarily translate to improved saf…

Math

Trusta: Reasoning about Assurance Cases with Formal Methods and Large Language Models

2023-09-22 · Zezhong Chen, Yuxin Deng, Wenjie Du

Assurance cases can be used to argue for the safety of products in safety engineering. In safety-critical areas, the construction of assurance cases is indispensable. Trustworthiness Derivation Trees (TDTs) enhance assur…

A Concrete Roadmap towards Safety Cases based on Chain-of-Thought Monitoring

2025-10-22 · Julian Schulz arxiv

As AI systems approach dangerous capability levels where inability safety cases become insufficient, we need alternative approaches to ensure safety. This paper presents a roadmap for constructing safety cases based on c…

Enhancing Model Defense Against Jailbreaks with Proactive Safety Reasoning

2025-01-31 · Xianglin Yang, Gelei Deng, Jieming Shi, Tianwei Zhang 외

Large language models (LLMs) are vital for a wide range of applications yet remain susceptible to jailbreak threats, which could lead to the generation of inappropriate responses. Conventional defenses, such as refusal a…

BlockingSafety Alignment