破解事实性知识：大语言模型中退化知识神经元的综合分析
Cracking Factual Knowledge: A Comprehensive Analysis of Degenerate Knowledge Neurons in Large Language Models
AI总结：
本文从结构和功能两方面严格定义退化知识神经元（DKNs），提出神经拓扑聚类方法以更准确地获取DKNs，并通过34项实验揭示DKNs与大语言模型鲁棒性、可演化性和复杂性之间的关联。
摘要：
大语言模型（LLMs）存储了大量事实性知识，但其底层机制仍不明确。先前研究表明，事实性知识存储在多层感知机权重中，且部分存储单元表现出退化性，被称为退化知识神经元（DKNs）。尽管这一概念具有新颖性和独特性质，但尚未被严格定义或系统研究。我们首先考察MLP神经元的连接权重模式，并从结构和功能两个方面定义DKNs。在此基础上，我们引入了神经拓扑聚类方法，该方法允许以任意数量和结构形成DKNs，从而实现更准确的DKN获取。此外，受认知科学启发，我们探索了DKNs与大语言模型的鲁棒性、可演化性和复杂性之间的关系。我们在6种设置下执行了34项实验，证明了DKNs与这三种性质之间的关联。代码即将公开。
英文摘要：
Large language models (LLMs) store extensive factual knowledge, but the underlying mechanisms remain unclear. Previous research suggests that factual knowledge is stored within multi-layer perceptron weights, and some storage units exhibit degeneracy, referred to as Degenerate Knowledge Neurons (DKNs). Despite the novelty and unique properties of this concept, it has not been rigorously defined or systematically studied. We first consider the connection weight patterns of MLP neurons and define DKNs from both structural and functional aspects. Based on this, we introduce the Neurological Topology Clustering method, which allows the formation of DKNs in any numbers and structures, leading to a more accurate DKN acquisition. Furthermore, inspired by cognitive science, we explore the relationship between DKNs and the robustness, evolvability, and complexity of LLMs. Our execution of 34 experiments under 6 settings demonstrates the connection between DKNs and these three properties. The code will be available soon.
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。
