基于大语言模型图式思维的个性化学习路径推荐算法
Personalized Learning Path Recommendation Algorithm Based on Schema Thinking of Large Language Models
刘克 1乔媛媛1
作者信息
- 1. 北京邮电大学人工智能学院，北京 100876
- 折叠
摘要
个性化学习路径规划通过分析学习者历史交互行为推荐有序的知识点序列，在提升在线教育效率与体验方面具有重要应用价值。然而，传统序列推荐方法（如SASRec、BERT4Rec）面临深层语义逻辑捕捉不足的问题，而直接应用大语言模型（LLM）则存在严重幻觉、长链路推理能力弱以及生成结果偏离教学大纲的局限。为此，提出一种基于检索增强图思维（RA-GoT）的个性化路径规划算法。该方法构建"相似路径检索 - 图式演化推理 - 候选约束映射"的结构化生成系统：利用 KNN 检索相似学习者的历史路径作为上下文参考，基于GoT框架执行"生成 - 批量对比打分 - 动态剪枝"的非线性规划，并通过模糊匹配机制将生成结果严格映射至知识库。在MOOCCubeX数据集上的实验表明，该方法的 Recall@10 和 NDCG@10 分别达到 65.81% 和 56.80%，相比 BERT4Rec 和 LLM-CoT等基线模型性能提升显著。该方法为大模型在教育推荐领域的落地提供了高可信、可解释的解决方案。
Abstract
Personalized learning path planning, which recommends ordered sequences of knowledge concepts by analyzing learners\' historical interactions, holds significant value for improving the efficiency of online education. However, traditional sequence recommendation methods (e.g., SASRec, BERT4Rec) struggle to capture deep semantic logic, while the direct application of Large Language Models (LLMs) suffers from severe hallucinations, weak long-term reasoning capabilities, and deviations from established syllabi. To this end, this paper proposes a Personalized Path Planning algorithm based on Retrieval-Augmented Graph of Thoughts (RA-GoT). This algorithm constructs a structured generation system of "Similar Path Retrieval - Graph-based Evolutionary Reasoning - Candidate Constraint Mapping": it utilizes KNN to retrieve historical paths of similar learners as context references, executes non-linear planning of "Generate - Contrastive Scoring - Dynamic Pruning" based on the GoT framework, and strictly maps generation results to the knowledge base via a fuzzy matching mechanism. Experiments on the MOOCCubeX dataset show that the proposed method achieves a Recall@10 of 65.81% and an NDCG@10 of 56.80%, demonstrating significant performance improvements over baselines such as BERT4Rec and LLM-CoT. This method provides a highly credible and explainable solution for the deployment of LLMs in educational recommendations.?????关键词
人工智能/个性化学习路径/思维图/大语言模型Key words
Artificial Intelligence/Personalized Learning Path/Graph of Thoughts/Large Language Model引用本文复制引用
刘克,乔媛媛.基于大语言模型图式思维的个性化学习路径推荐算法[EB/OL].(2026-02-09)[2026-02-11].http://www.paper.edu.cn/releasepaper/content/202602-50.学科分类
计算技术、计算机技术
评论