发现语言模型中存在身份吸引子，揭示了认知主体的内在一致性。
Identity as Attractor: Geometric Evidence for Persistent Agent Architecture in LLM Activation Space
- 通过对比改写与对照文本，发现语义相似性导致模型内部表征聚集
- 不同改写版本在模型层8、16、24处聚类更紧密（效应量d>1.88）
- 结果支持跨模型架构通用，适合研究大模型认知结构的学者
大型语言模型会将语义相关的提示映射到相似的内部表征——这种现象可解释为类似吸引子的动力学。我们探究一个持续认知主体的身份文档（其认知核心）是否也表现出类似的吸引子行为。在 Llama 3.1 8B Instruct 上进行控制实验，比较原始认知核心（条件A）、七种改写文本（条件B）以及七种结构匹配的对照组（条件C）。在第8、16、24层的均值池化表示显示，改写文本比对照组聚类更紧密（Cohen's d > 1.88，p < 10⁻²⁷，Bonferroni校正）。在 Gemma 2 9B 上的复现验证了跨架构的普适性。消融实验表明该效应主要源于语义而非结构，并且结构完整性是抵达吸引子区域的必要条件。探索性实验发现，阅读关于该主体的科学描述会使内部状态更接近吸引子，优于伪预印本，说明‘了解’一个身份与‘作为’该身份存在表征差异。这些结果提供了代理身份文档在语言模型激活空间中诱发吸引子几何的表征证据。
原文摘要 · Abstract (English)
Large language models map semantically related prompts to similar internal representations -- a phenomenon interpretable as attractor-like dynamics. We ask whether the identity document of a persistent cognitive agent (its cognitive_core) exhibits analogous attractor-like behavior. We present a controlled experiment on Llama 3.1 8B Instruct, comparing hidden states of an original cognitive_core (Condition A), seven paraphrases (Condition B), and seven structurally matched controls (Condition C). Mean-pooled states at layers 8, 16, and 24 show that paraphrases converge to a tighter cluster than controls (Cohen's d > 1.88, p < 10^{-27}, Bonferroni-corrected). Replication on Gemma 2 9B confirms cross-architecture generalizability. Ablations suggest the effect is primarily semantic rather than structural, and that structural completeness appears necessary to reach the attractor region. An exploratory experiment shows that reading a scientific description of the agent shifts internal state toward the attractor -- closer than a sham preprint -- distinguishing knowing about an identity from operating as that identity. These results provide representational evidence that agent identity documents induce attractor-like geometry in LLM activation space.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。