Reproduce “Towards LLM-Empowered Knowledge Tracing via LLM-Student Hierarchical Behavior Alignment in Hyperbolic Space” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Xingcheng Fu, Shengpeng Wang, Yisen Gao, Xianxian Li, Chunpei Li, Qingyun Sun, Dongran Yu

The L-HAKT framework introduces large language models to simulate student learning behaviors and construct hierarchical knowledge graphs, integrating these with hyperbolic geometry for tracking student concept mastery. This approach yielded superior performance on four real-world datasets, consistently surpassing state-of-the-art baselines with improvements of up to 0.84% AUC and 0.56% accuracy.

Key insights

  • Hyperbolic space, with its inherent negative curvature, is uniquely suited to model the tree-like, hierarchical structure of knowledge points and their relationships, offering a geometric advantage over Euclidean embeddings.
  • Large Language Models can function as "teacher" agents to semantically parse questions, build fine-grained hierarchical knowledge graphs, and quantify exercise difficulty, providing rich pedagogical structure.
  • LLMs can also act as "student" agents to simulate dynamic cognitive processes, such as engagement and hierarchical forgetting, effectively augmenting sparse student interaction data with plausible "thinking paths."

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.