arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.06650cs.CL
收藏

超越静态对齐:通过风险感知的思维链实现LLM安全的分层策略控制

Beyond Static Alignment: Hierarchical Policy Control for LLM Safety via Risk-Aware Chain-of-Thought

Qiyuan Tech, Beijing, China(奇渊科技,北京,中国)

Jianfeng Si, Lin Sun, Weihong Lin, Xiangzheng Zhang

AI总结:

PACT通过分层策略架构和风险感知推理,解决LLM安全与有用性之间的权衡问题,提升动态安全控制和可控性。

摘要:

大型语言模型(LLMs)由于静态的一刀切安全策略缺乏运行时可控性,面临安全与有用性之间的根本性权衡,这使得难以根据多样化应用需求定制响应。结果,模型可能对善意请求过度拒绝或对有害请求约束不足。我们提出了PACT(通过思维链配置的操作),一种通过显式、风险感知的推理实现动态安全控制的框架。PACT采用分层策略架构:一个不可覆盖的全球安全策略为关键风险(如儿童安全、暴力极端主义)建立不可更改的边界,而用户定义的策略可以引入领域特定(非全球)风险类别,并指定标签到操作行为以提高实际部署中的实用性。该框架将安全决策分解为结构化的分类→操作路径,将查询路由到适当的操作(遵守、引导或拒绝),并使决策过程透明。大量实验表明,PACT在全局政策评估下实现了接近最先进的安全性能,同时在用户特定政策评估下实现了最佳可控性,有效缓解了安全与有用性之间的权衡。我们将发布PACT模型套件、训练数据和评估协议,以促进可控安全对齐的可重复研究。

英文摘要:

Large Language Models (LLMs) face a fundamental safety-helpfulness trade-off due to static, one-size-fits-all safety policies that lack runtime controllabilityxf, making it difficult to tailor responses to diverse application needs. %As a result, models may over-refuse benign requests or under-constrain harmful ones. We present \textbf{PACT} (Prompt-configured Action via Chain-of-Thought), a framework for dynamic safety control through explicit, risk-aware reasoning. PACT operates under a hierarchical policy architecture: a non-overridable global safety policy establishes immutable boundaries for critical risks (e.g., child safety, violent extremism), while user-defined policies can introduce domain-specific (non-global) risk categories and specify label-to-action behaviors to improve utility in real-world deployment settings. The framework decomposes safety decisions into structured Classify$\rightarrow$Act paths that route queries to the appropriate action (comply, guide, or reject) and render the decision-making process transparent. Extensive experiments demonstrate that PACT achieves near state-of-the-art safety performance under global policy evaluation while attaining the best controllability under user-specific policy evaluation, effectively mitigating the safety-helpfulness trade-off. We will release the PACT model suite, training data, and evaluation protocols to facilitate reproducible research in controllable safety alignment.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑