Project Ariadne: 一种用于审计大语言模型代理忠实性的结构因果框架
Project Ariadne: A Structural Causal Framework for Auditing Faithfulness in LLM Agents
AI总结：
Project Ariadne提出一种结构因果框架，通过因果干预评估大语言模型代理推理的忠实性，揭示代理推理与决策之间的因果脱节问题。
摘要：
随着大语言模型（LLM）代理被越来越多地用于高风险的自主决策，其推理过程的透明性已成为关键的安全问题。虽然链式思维（CoT）提示允许代理生成可读的推理轨迹，但这些轨迹是否是模型输出的忠实生成驱动因素还是仅仅是事后解释仍不明确。我们引入了Project Ariadne，一种新的XAI框架，利用结构因果模型（SCMs）和反事实逻辑来审计代理推理的因果完整性。与现有可解释性方法仅依赖表面文本相似性不同，Project Ariadne对中间推理节点执行硬干预（do- calculus），系统性地反转逻辑、否定前提和逆转事实主张，以测量终端答案的因果敏感性（ϕ）。我们对最先进的模型进行实证评估，发现存在持续的忠实性差距。我们定义并检测了一种广泛存在的故障模式，称为因果解耦，其中代理在事实和科学领域中违反密度（ρ）高达0.77。在这些情况下，代理得出相同结论，尽管内部逻辑矛盾，证明其推理轨迹作为“推理剧院”而决策由潜在参数先验控制。我们的发现表明，当前代理架构本质上容易产生不忠实的解释，我们提出Ariadne分数作为新的基准，以对齐声明的逻辑与模型行为。
英文摘要：
As Large Language Model (LLM) agents are increasingly tasked with high-stakes autonomous decision-making, the transparency of their reasoning processes has become a critical safety concern. While \textit{Chain-of-Thought} (CoT) prompting allows agents to generate human-readable reasoning traces, it remains unclear whether these traces are \textbf{faithful} generative drivers of the model's output or merely \textbf{post-hoc rationalizations}. We introduce \textbf{Project Ariadne}, a novel XAI framework that utilizes Structural Causal Models (SCMs) and counterfactual logic to audit the causal integrity of agentic reasoning. Unlike existing interpretability methods that rely on surface-level textual similarity, Project Ariadne performs \textbf{hard interventions} ($do$-calculus) on intermediate reasoning nodes -- systematically inverting logic, negating premises, and reversing factual claims -- to measure the \textbf{Causal Sensitivity} ($ϕ$) of the terminal answer. Our empirical evaluation of state-of-the-art models reveals a persistent \textit{Faithfulness Gap}. We define and detect a widespread failure mode termed \textbf{Causal Decoupling}, where agents exhibit a violation density ($ρ$) of up to $0.77$ in factual and scientific domains. In these instances, agents arrive at identical conclusions despite contradictory internal logic, proving that their reasoning traces function as "Reasoning Theater" while decision-making is governed by latent parametric priors. Our findings suggest that current agentic architectures are inherently prone to unfaithful explanation, and we propose the Ariadne Score as a new benchmark for aligning stated logic with model action.
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。
