AlignKT: Explicitly Modeling Knowledge State for Knowledge Tracing with Ideal State Alignment

This is the code for the paper AlignKT: Explicitly Modeling Knowledge State for Knowledge Tracing with Ideal State Alignment. See our paper here(the link will be updated here).

Requirements

torch>=2.1.0

Arguments

• --dataset_name: default value=assist2009, value={algebra2005, nips_task34}
• --d_model: default value=384(for AS09), value={256(for AL05, NIPS34)}
• --d_ff: default value=1024(for AS09, NIPS34), value={768(for AL05)}
• --n_heads: default value=4(for AS09), value={8(for AL05, NIPS34)}
• --drop_out: default value=0.25(for AS09), value={0.2(for AL05, NIPS34)}
• --batch_size: default value=128(for AS09), value={32(for AL05, NIPS34)}

Run

python main.py --seed 42

Citation

The citation information for the paper will be updated here.

Acknowledgements

The code for model training and real-scenario evaluation is sourced from pykt-team/pykt-toolkit: pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models