COMPASS：基于上下文调节的PID注意力转向系统用于减少幻觉
COMPASS: Context-Modulated PID Attention Steering System for Hallucination Mitigation
AI总结：
COMPASS通过上下文调节PID注意力转向系统减少大型语言模型的幻觉，通过可解释的反馈回路提升生成的准确性与可解释性。
摘要：
大型语言模型（LLMs）尽管能够访问相关证据，仍然常常生成流畅但事实错误的陈述，这种失败模式源于它们在上下文知识和参数知识之间分配注意力的方式。理解并引导这种内部行为对于可靠部署和模型机制的科学可解释性至关重要。我们介绍了COMPASS（上下文调节PID注意力转向系统），一种轻量、可解释的控制框架，它在解码过程中直接嵌入了一个基于模型的反馈回路。COMPASS通过一个透明的指标，即上下文依赖性分数（CRS），量化上下文依赖性，该指标作为在线探测注意力头如何将生成扎根于证据的工具。利用这个可解释的信号，PID控制器动态调节注意力头，以在不重新训练或多次解码的情况下保持事实一致性。在多个基准测试（HotpotQA、XSum、HaluEval、RAGTruth）中，COMPASS一致地降低了上下文幻觉率（绝对值减少2.8到5.8个百分点），并揭示了不同的注意力头如何贡献于证据对齐。这些结果突显了反馈驱动的可解释性作为理解LLM行为科学途径的重要性。
英文摘要：
Large language models (LLMs) often generate fluent but factually incorrect statements despite having access to relevant evidence, a failure mode rooted in how they allocate attention between contextual and parametric knowledge. Understanding and steering this internal behavior is key both for trustworthy deployment and for scientific interpretability of model mechanisms. We introduce COMPASS (Context-Modulated PID Attention Steering System), a lightweight, interpretable control framework that embeds a model-based feedback loop directly within decoding. COMPASS quantifies context reliance via a transparent metric, the Context Reliance Score (CRS), which serves as an online probe of how attention heads ground generation in evidence. Using this interpretable signal, a PID controller dynamically modulates attention heads to maintain factual consistency without retraining or multi-pass decoding. Across benchmarks (HotpotQA, XSum, HaluEval, RAGTruth), COMPASS consistently reduces contextual hallucination rates (2.8 to 5.8 percent absolute) while revealing how distinct attention heads contribute to evidence alignment. These results highlight feedback-driven interpretability as a pathway toward scientific understanding of LLM behavior.
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。
