Compound AI System for Personalized Learning: Integrating LLM Agents with Knowledge Graphs

Personalized learning plays a critical role in enhancing educational outcomes by customizing experiences to meet individual learners' unique needs, preferences, and abilities. Although Large Language Models (LLMs) offer promising opportunities for personalization through natural language interactions, they face challenges related to maintaining long-term context and ensuring traceability of a learner's evolving knowledge. This paper introduces a composite AI framework that merges an LLM-powered conversational agent with Knowledge Graphs (KGs) structured according to Knowledge Space Theory (KST), enhanced by advanced retrieval tools and a robust long-term memory mechanism. This system autonomously orchestrates meaningful, context-aware dialogues while utilizing function calling to access various tools, effectively bridging the gap between dynamic interactions and structured knowledge representation. Our empirical evaluations indicate that the proposed system excels in maintaining coherent dialogues, accurately tracking learner progress, and providing reliable, personalized educational support. By addressing the limitations of standalone LLMs and traditional adaptive learning systems, this integrated approach signifies a notable advancement in intelligent educational technology.