arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.05097cs.LGcs.AIcs.CL
收藏

连续知识更新在大语言模型系统中:通过多时间尺度记忆动态学习

Continual Knowledge Updating in LLM Systems: Learning Through Multi-Timescale Memory Dynamics

The Cyprus Institute, Nicosia, Cyprus(塞浦路斯研究所,尼科西亚,塞浦路斯)

Andreas Pattichis, Constantine Dovrolis

AI总结:

本文提出基于多时间尺度动态的记忆机制,通过耦合内部变量实现知识的持续更新与学习,重构外部记忆作为学习的基础 substrates。

摘要:

大语言模型在训练后被部署到一个不断变化的世界中,外部记忆对此进行补偿,但大多数系统通过显式管理而非让其自行适应。生物记忆的工作方式不同:耦合的多时间尺度动态使新关联立即可用,强化通过重复确认的内容,并让其余部分逐渐遗忘。我们主张外部记忆应遵循类似原则。在Memini中,这一观点表现为一种关联记忆,将知识组织为有向图。每条边携带两个耦合的内部变量,一个快一个慢,遵循Benna-Fusi模型的突触巩固机制。从这种耦合中,事件敏感性、渐进巩固和选择性遗忘作为单一机制的特征出现,重新将外部记忆视为一种学习基质,通过自身动态重新组织。

英文摘要:

LLMs are trained once, then deployed into a world that never stops changing. External memory compensates for this, but most systems manage it explicitly rather than letting it adapt on its own. Biological memory works differently: coupled multi-timescale dynamics make new associations immediately usable, strengthen what repetition confirms, and let the rest fade. We argue that external memory should follow a similar principle. In Memini, this view takes the form of an associative memory that organizes knowledge as a directed graph. Each edge carries two coupled internal variables, one fast and one slow, following the Benna-Fusi model of synaptic consolidation. From this coupling, episodic sensitivity, gradual consolidation, and selective forgetting are expected to emerge as facets of a single mechanism, reframing external memory as a learning substrate that reorganizes through its own dynamics. This workshop article describes an early-stage conceptual design without experimental evaluation.

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。

↑