Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.15397 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914398397792256 |
|---|---|
| author | Pan, Yu Yu, Wenlong Wu, Tiejun Ye, Xiaohu Si, Qiannan Xu, Guangquan Wu, Bin |
| author_facet | Pan, Yu Yu, Wenlong Wu, Tiejun Ye, Xiaohu Si, Qiannan Xu, Guangquan Wu, Bin |
| contents | Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks. However, they remain highly susceptible to jailbreak attacks that undermine their safety alignment. Existing defense mechanisms typically rely on post hoc filtering applied only to the final output, leaving intermediate reasoning steps unmonitored and vulnerable to adversarial manipulation. To address this gap, this paper proposes a SaFer Chain-of-Thought (SFCoT) framework, which proactively evaluates and calibrates potentially unsafe reasoning steps in real time. SFCoT incorporates a three-tier safety scoring system alongside a multi-perspective consistency verification mechanism, designed to detect potential risks throughout the reasoning process. A dynamic intervention module subsequently performs targeted calibration to redirect reasoning trajectories toward safe outcomes. Experimental results demonstrate that SFCoT reduces the attack success rate from $58.97\%$ to $12.31\%$, demonstrating it as an effective and efficient LLM safety enhancement method without a significant decline in general performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_15397 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SFCoT: Safer Chain-of-Thought via Active Safety Evaluation and Calibration Pan, Yu Yu, Wenlong Wu, Tiejun Ye, Xiaohu Si, Qiannan Xu, Guangquan Wu, Bin Cryptography and Security Artificial Intelligence Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks. However, they remain highly susceptible to jailbreak attacks that undermine their safety alignment. Existing defense mechanisms typically rely on post hoc filtering applied only to the final output, leaving intermediate reasoning steps unmonitored and vulnerable to adversarial manipulation. To address this gap, this paper proposes a SaFer Chain-of-Thought (SFCoT) framework, which proactively evaluates and calibrates potentially unsafe reasoning steps in real time. SFCoT incorporates a three-tier safety scoring system alongside a multi-perspective consistency verification mechanism, designed to detect potential risks throughout the reasoning process. A dynamic intervention module subsequently performs targeted calibration to redirect reasoning trajectories toward safe outcomes. Experimental results demonstrate that SFCoT reduces the attack success rate from $58.97\%$ to $12.31\%$, demonstrating it as an effective and efficient LLM safety enhancement method without a significant decline in general performance. |
| title | SFCoT: Safer Chain-of-Thought via Active Safety Evaluation and Calibration |
| topic | Cryptography and Security Artificial Intelligence |
| url | https://arxiv.org/abs/2603.15397 |