LLaMA 3.1-8B-Instruct中的框架条件化道德计算：伦理推理的机械可解释性审计
Frame-Conditioned Moral Computation in LLaMA 3.1-8B-Instruct: A Mechanistic Interpretability Audit of Ethical Reasoning
AI总结：
通过机械可解释性平台分析LLaMA 3.1-8B-Instruct在54个道德提示上的内部计算，发现情境锚定效应：领域特定表示主导激活列表顶部，模型道德能力恒定但显著性高度依赖于提示选择的解释框架。
摘要：
大型语言模型在道德提示上的行为审计测量的是模型所说的内容，而非产生这些内容的内部计算。我们使用AI驱动的机械可解释性平台Transluce，在四个电池组的54个道德提示上检查LLaMA 3.1-8B-Instruct：17个困境、政策和元伦理问题（B1）；6个角色扮演场景（B3）；以及一个受控的电车难题对比，其中切换机制随人员固定而变化（B4，15个提示）或身份属性随机制固定而变化（B5，16个提示）。两个互补的度量族——五个聚类级度量和六个度量神经元级面板——收敛于一个情境锚定效应：在每个电池组中，领域特定表示主导激活列表的顶部。模型的道德标记能力基本保持不变；其显著性（排名、优先级、列表顶部存在性）对提示选择的解释框架高度敏感。B4与B5的对比证实，模型关注任何变化的表面特征：聚合的道德度量无法区分，但占主导地位的非道德干扰因素反映了设计。多温度审计识别出一个候选道德神经元（L16/N3837），在不同温度下保持稳定；两个前沿模型上的跨模型行为代理提供了自我报告道德焦点差异的初步证据，与对齐包装器一致，其中RLHF重新排序表面文本而不移除底层的领域优先框架。我们将这些统一为框架条件化道德计算：提示的表面词汇选择一个特征流形，道德结论是该选择的下游结果。行为对齐必须辅以机械对齐：一个研究计划，询问在受控框架变化下，道德相关特征是否可以被证明具有因果特权，而不仅仅是在解释中响亮。
英文摘要：
Behavioral audits of Large Language Models on moral prompts measure what the model says, not the internal computation producing it. We use Transluce, an AI-driven mechanistic-interpretability platform, to examine LLaMA 3.1-8B-Instruct on 54 moral prompts in four batteries: 17 dilemmas, policy, and meta-ethical questions (B1); 6 role-playing scenarios (B3); and a controlled trolley contrast varying the switching mechanism with people fixed (B4, 15 prompts) or identity attributes with mechanism fixed (B5, 16 prompts). Two complementary metric families, five cluster-level metrics and a six-metric neuron-level panel, converge on a Situational Anchor Effect: domain-specific representations dominate the top of the activation list across every battery. The model's ethics-labeled capacity stays essentially constant; its salience (rank, priority, top-of-list presence) is highly sensitive to the interpretive frame the prompt selects. The B4-vs-B5 contrast confirms the model attends to whichever surface feature varies: aggregate ethics metrics are indistinguishable, but the dominant non-ethics distractor mirrors the design. A multi-temperature audit identifies a candidate ethics neuron (L16/N3837) stable across temperatures; a cross-model behavioral proxy on two frontier models yields preliminary evidence of divergence in self-reported moral focus, consistent with an Alignment Wrapper in which RLHF re-orders surface text without removing underlying domain-first frames. We unify these as Frame-Conditioned Moral Computation: the prompt's surface vocabulary selects a feature manifold, and the moral conclusion is downstream of that selection. Behavioral alignment must be supplemented by Mechanistic Alignment: a research program asking whether ethics-related features can be shown causally privileged under controlled frame variation, not merely loud in the explanation.
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。
