UniKT¶
Knowledge tracing (KT) estimates students’ knowledge states from their historical response sequences and predicts whether they will answer the next question correctly. UniKT trains and evaluates knowledge tracing models under a unified data pipeline, student-level cross-validation splits, and standardized metric definitions, enabling fair comparison and complete reproducibility across different models.
65 个模型实现
14 Public Datasets
150M+ Learning Interactions
1.2M+ Students
Documentation Navigation¶
Quick Start Configure the Pixi environment and complete your first data processing and training run.
Model Zoo Browse papers, author implementations, and integration notes by model.
Datasets Explore data sources, scales, schemas, and processing pipelines.
Evaluation Benchmark leaderboard, cross-validation protocol, standalone testing, case analysis, and efficiency benchmarking.
User Guides Training, hyperparameter search, experiment tracking, and Web platform.
Developer Guides Adding new models, custom datasets, multi-stage training, and callbacks.
API Reference Command-line arguments and Python interfaces.
Changelog Framework and documentation release notes.