KT²: Knowledge-Tree-based Knowledge Tracing

This repository contains the official implementation of the paper:

A Hierarchical Probabilistic Framework for Incremental Knowledge Tracing in Classroom Settings

This framework introduces a probabilistic approach that models student knowledge over a structured KC tree, with support for real-time updates via expectation-maximization (EM) and graph-based propagation. The method is designed for classroom scenarios and emphasizes interpretability, modularity, and incremental updates.

Project Structure

KT2/
├── scripts/             # Entry scripts for graph init and training
├── KT2/                 # Core algorithm modules
├── KC_tree/             # Knowledge concept tree construction and pruning
├── config/              # Configuration files and defaults
├── data/                # Datasets and KC trees
├── output/              # Output directory
├── preprocess/          # Data preprocessing tools
└── requirements.txt     # Python dependencies

Installation

To install dependencies:

conda create -n kt2 python=3.10
conda activate kt2
pip install -r requirements.txt

Quick Start

1. Build the Knowledge Graph

To build and prune the KC graph for a given dataset:

python scripts/init_graph.py --dataset DATASET_NAME

This will load data from data/dataset/DATASET_NAME/ and save the KC tree template to data/KC_tree/DATASET_NAME/.

2. Run the KT² Algorithm

Once the graph is ready, run the full knowledge tracing pipeline:

python scripts/run_KT2.py

Paths and parameters can be configured in config/config.yaml. To run the experiment of each classroom setting, set dataset as XES3G5M or MOOCRadar. Set root_node as Application_Module/Computation_Module/Counting_Module for XES3G5M; Wine_Knowledge/Circuit_Design/Education_Theory for MOOCRadar.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Citation

If you find this project helpful or use it in your research, please consider citing:

@misc{gao2025kt2,
      title={A Hierarchical Probabilistic Framework for Incremental Knowledge Tracing in Classroom Settings}, 
      author={Xinyi Gao and Qiucheng Wu and Yang Zhang and Xuechen Liu and Kaizhi Qian and Ying Xu and Shiyu Chang},
      year={2025},
      eprint={2506.09393},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2506.09393}, 
}