TSDR is robust to the smoothness coefficient, with a stable sweet spot for lambda in [0.3, 1.0]¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? On the Prob dataset, temporal smoothness outperforms vanilla DR (lambda = 0) within an appropriate range, with a stable sweet spot generally in lambda in [0.3, 1.0]. → Peilin Zhan 2026

Evidence¶

Peilin Zhan 2026¶

Peilin Zhan, Wei Chen, Weilin Chen, Shuyi Pan, Ruichu Cai. (2026). Temporal Smoothness Doubly Robust Learning for Debiased Knowledge Tracing. https://arxiv.org/abs/2605.05958 doi:10.48550/arxiv.2605.05958

q2 · i? · design · r2

Sensitivity analysis (RQ3, Figure 4) varying lambda over {0, 0.1, 0.3, 0.5, 0.7, 1, 2} on the Prob dataset across six backbones, measuring AUC and ACC. The article reports "a stable “sweet spot” generally inλ∈[0.3,1.0]."

"The method exhibits strong robustness, achieving a stable “sweet spot” generally inλ∈[0.3,1.0]."

Discussion¶

Related Claims¶

• TSDR's AUC gains are largest on open-ended platforms and limited on EdNet, whose mandatory bundle completion policy constrains selection bias — related
• TSDR improves the AUC of six state-of-the-art KT backbones across nine real-world benchmark datasets — a broader claim this one bears on
• In ablations, temporal smoothness works best on the imputation branch, and the full TSDR variant achieves the best AUC and ACC on all six backbones — related
• Under controlled synthetic MNAR bias, TSDR-equipped models improve across the entire bias spectrum, with the RMSE advantage diminishing beyond gamma > 0.8 — related