隐喻追踪器：基于理论指导的隐状态分析
Metaphor Tracer: A Theory-Informed Analysis of Hidden States
浏览论文内容
中文总结 AI 辅助
本研究提出Metaphor Tracer工具，通过聚合器与区分器分析语言模型隐状态，验证其可追踪文本结构与隐喻迁移，为隐状态提供关系论解读。
中文摘要 AI 辅助
语言模型的隐状态能揭示单篇文本的组织方式吗？无需训练，仅通过一次前向传播，我们就能对每个词元位置的两个属性进行评分：聚合器（aggregator）衡量该位置是否将整篇文本整合为稳定构型；区分器（differentiator）衡量模型阅读时，其他词元是否会短暂进入该位置的子空间，即隐喻的本义——迁移。我们在一篇发现文本上冻结了常量，其余所有实验均为验证性。聚合器并非经典意义上的信息度量或显著性度量。在三个不相关模型中，当某个能指重复出现时，其意外值（surprisal）和注意力会衰减，但聚合器评分保持稳定：该通道标记了词元在文本中的位置。这一结果基于独立的真实依据：聚合器可追踪人工设计的语域边界（6/6个单元），以及心理分析师在工具诞生前对临床文本的标注（34/36个单元），其评分在词汇对照之上呈分级递增，且存在类型级度量无法复现的解离现象。迁移测试验证了结果的形态：词元结构随词汇类型迁移的模型，对单篇论述的阅读效果最差；在匹配的基础模型/指令微调模型对中，微调提升了保真度但未改变类型迁移。结构价值是词元在当前文本中位置的属性，而非仅其向量的属性：这是对隐状态的关系论而非本质论解读，实现了早于该工具的理论。
英文摘要
What do a language model's hidden states say about the organization of a single text? From one forward pass, without training, we score every token position on two properties. The *aggregator* measures whether the position consolidates the whole text into a stable configuration. The *differentiator*, whether other tokens are transiently carried into its subspace as the model reads: metaphor in its root sense, transport. Constants were frozen on one discovery text; every other is confirmatory. The aggregator is not, in the classic sense, an information measure, nor a measure of salience. Across three unrelated models, as a signifier repeats, its surprisal and its attention drain while its aggregator score holds: the channel marks a token's place in the text. That this tracks a reading rests on independent ground truth: an engineered register the aggregator follows across its boundaries (6/6 cells), and a psychoanalyst's marking of clinical transcripts, fixed before the instrument existed, in 34/36 cells, with a graded increment above lexical controls and dissociations no type-level measure reproduces. A transfer test gives the result its shape: the model whose token structure travels with lexical type reads the singular discourse worst, and in a matched base/instruct pair tuning raises fidelity without moving type-transfer. Structural value is a property of a token's place in *this* text, not of its vector alone: a relational rather than essentialist reading of hidden states, operationalizing theory that predated the instrument.
发表机构
- Hermeneutic AI（诠释学人工智能公司）
机构由 AI 辅助整理，请以论文原文为准。
