Research

Paper

AI LLM March 16, 2026

SFCoT: Safer Chain-of-Thought via Active Safety Evaluation and Calibration

Authors

Yu Pan, Wenlong Yu, Tiejun Wu, Xiaohu Ye, Qiannan Si, Guangquan Xu, Bin Wu

Abstract

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.

Metadata

arXiv ID: 2603.15397
Provider: ARXIV
Primary Category: cs.CR
Published: 2026-03-16
Fetched: 2026-03-17 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.15397v1</id>\n    <title>SFCoT: Safer Chain-of-Thought via Active Safety Evaluation and Calibration</title>\n    <updated>2026-03-16T15:13:21Z</updated>\n    <link href='https://arxiv.org/abs/2603.15397v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.15397v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>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.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CR'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-16T15:13:21Z</published>\n    <arxiv:primary_category term='cs.CR'/>\n    <author>\n      <name>Yu Pan</name>\n    </author>\n    <author>\n      <name>Wenlong Yu</name>\n    </author>\n    <author>\n      <name>Tiejun Wu</name>\n    </author>\n    <author>\n      <name>Xiaohu Ye</name>\n    </author>\n    <author>\n      <name>Qiannan Si</name>\n    </author>\n    <author>\n      <name>Guangquan Xu</name>\n    </author>\n    <author>\n      <name>Bin Wu</name>\n    </author>\n  </entry>"
}