---
source_url: https://arxiv.org/abs/2605.05958
ingested: 2026-05-13
sha256: 5c5cdc1d1e07eb3952d4991685e8e0d45ef6496a04bed1a56a9ec9d53109319d
---

# Temporal Smoothness Doubly Robust Learning for Debiased Knowledge Tracing
arXiv: 2605.05958
Authors: Zhan, P., Chen, W., Chen, W., Pan, S., Cai, R.
Published: 2026-05-07
Categories: cs.AI
Venue: arXiv preprint

## Abstract
KT systems suffer from selection bias. This paper introduces TSDR, a doubly robust framework with temporal smoothness regularization that jointly optimizes the KT predictor and imputation model, reducing variance while preserving unbiasedness guarantees.