Adversarial contrastive learning to augment deep knowledge tracing

doi:10.1016/j.engappai.2026.115001

Adversarial contrastive learning to augment deep knowledge tracing

2026-05-08

0

PRE

AI

DOI：10.1016/j.engappai.2026.115001

Original

Original request for help

Save

Abstract

En 中文

• To address KT overfitting and low sample distinguishability in sparse data, we propose the ACDKT model. • ACDKT integrates a multi-knowledge attention mechanism to enrich question representation and improve contrastive learning. • Experiments on real-world educational datasets show ACDKT improves accuracy, reduces overfitting, and generalizes well.

Keywords:

ACDKT

knowledge tracing

contrastive learning

adversarial learning

sparse data

Adversarial contrastive learning to augment deep knowledge tracing

Abstract

Journal

Organization