Reproduce “SINKT: A Structure-Aware Inductive Knowledge Tracing Model with Large Language Model” 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
Lingyue Fu, Hao Guan, Kounianhua Du, Jianghao Lin, Wei Xia, Weinan Zhang, Ruiming Tang, Yasheng Wang and 1 more
SINKT introduces an inductive knowledge tracing model that integrates Large Language Models (LLMs) for generating explicit structural knowledge and semantic embeddings, combined with Graph Neural Networks (GNNs). This approach enables accurate predictions for unseen questions, achieving state-of-the-art performance and addressing the cold start problem in intelligent tutoring systems.
Key insights
- Large Language Models (LLMs) can effectively generate high-quality, rational structural knowledge, such as concept prerequisites, which significantly enhances knowledge tracing models.
- The explicit integration of both rich semantic information from content texts and intricate structural relationships from a knowledge graph is crucial for realizing inductive knowledge tracing.
- A multi-component architecture combining semantic encoders, graph neural networks for structural reasoning, and recurrent networks for student state modeling provides a robust framework for adaptive and scalable educational systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.