SALT: 通过泄漏自由思考引导激活以避免链式思维中的信息泄露
SALT: Steering Activations towards Leakage-free Thinking in Chain of Thought
AI总结：
SALT通过在链式思维中注入引导向量减少隐私泄露，提升模型隐私保护能力
摘要：
随着大型语言模型（LLMs）发展为能够访问敏感用户数据的个人助手，它们面临一个关键的隐私挑战：尽管先前的工作已经解决了输出层面的隐私问题，但最近的研究发现，LLMs经常通过其内部推理过程泄露私人信息，违反了上下文隐私期望。这些泄露的想法发生在模型无意中在推理轨迹中暴露敏感细节时，即使最终输出看起来是安全的。挑战在于在不损害模型推理能力的情况下防止此类泄漏，需要在隐私和效用之间寻求微妙的平衡。我们引入了Steering Activations towards Leakage-free Thinking（SALT），这是一种轻量级的测试时间干预方法，通过在隐藏状态中注入针对性的引导向量来减轻模型链式思维（CoT）中的隐私泄漏。我们识别出负责此行为的高泄漏层。通过在多个LLMs上进行实验，我们证明SALT在上下文隐私泄漏数据集AirGapAgent-R上实现了CPL的减少，包括在QwQ-32B上减少了18.2%，在Llama-3.1-8B上减少了17.9%，在Deepseek上减少了31.2%。同时保持了可比的任务性能和效用。我们的工作为在推理能力语言模型中建立SALT作为测试时间隐私保护的实用方法，提供了通往更安全部署LLM个人代理的路径。
英文摘要：
As Large Language Models (LLMs) evolve into personal assistants with access to sensitive user data, they face a critical privacy challenge: while prior work has addressed output-level privacy, recent findings reveal that LLMs often leak private information through their internal reasoning processes, violating contextual privacy expectations. These leaky thoughts occur when models inadvertently expose sensitive details in their reasoning traces, even when final outputs appear safe. The challenge lies in preventing such leakage without compromising the model's reasoning capabilities, requiring a delicate balance between privacy and utility. We introduce Steering Activations towards Leakage-free Thinking (SALT), a lightweight test-time intervention that mitigates privacy leakage in model's Chain of Thought (CoT) by injecting targeted steering vectors into hidden state. We identify the high-leakage layers responsible for this behavior. Through experiments across multiple LLMs, we demonstrate that SALT achieves reductions including $18.2\%$ reduction in CPL on QwQ-32B, $17.9\%$ reduction in CPL on Llama-3.1-8B, and $31.2\%$ reduction in CPL on Deepseek in contextual privacy leakage dataset AirGapAgent-R while maintaining comparable task performance and utility. Our work establishes SALT as a practical approach for test-time privacy protection in reasoning-capable language models, offering a path toward safer deployment of LLM-based personal agents.
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。
