September 20, 2025

Denoised Attention and Question-Augmented Representations for Knowledge Tracing

View Full Paper

Ask AI

Bookmark

DenoiseKT effectively solves the attention noise problem in knowledge tracing, improving prediction accuracy for student performance.
Experimental results demonstrate that DenoiseKT outperforms 22 state-of-the-art knowledge tracing models on four public datasets.
The model utilizes a graph neural network to refine input feature representations by capturing complex relationships between questions.
By introducing a denoised attention mechanism, DenoiseKT reduces attention weight on irrelevant information, enhancing overall performance.

Discussion

No takes yet. Share an insight, caveat, or question.

Novel model DenoiseKT addresses attention noise in knowledge tracing, highlighting its effectiveness and relevance.

Deng et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark