EVOL variants remain the strongest methods when recommended paths are executed on alternative DKT simulator instances¶

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? Policies trained against the original DKT (seed 42) and executed on four alternative DKT instances keep EVOL variants strongest, with graph-conditioned baselines degrading more sharply. → Geonwoo Bang 2026

Evidence¶

Geonwoo Bang 2026¶

Geonwoo Bang, Dongho Kim, and Moohong Min. (2026). EVOL: Simulator-Guided Evolutionary Expert Synthesis for Deployment-Free Learning Path Recommendation. arXiv:2610.03273. https://arxiv.org/abs/2610.03273 doi:10.48550/arxiv.2610.03273

q2 · i? · design · r2

Simulator-instance shift stress test (Table 6, ASSIST15 L=10): policies trained on the seed-42 DKT executed on four alternative DKTs (seeds 7, 100, 200, 333; validation AUC 0.700–0.708 vs 0.704). EVOL-BC cross-instance mean +0.341 vs PPO-vanilla +0.250.

"First, all three EVOL variants remain the strongest methods across the alternative DKT instances, and their differences are much smaller than the gap to the strongest baseline."

Discussion¶

Related Claims¶

• Under deployment-free inference, at least one EVOL variant achieves the best test EP in every dataset–length setting across three datasets and eight baselines — related
• EVOL variants remain the strongest methods under stochastic response models, weakening the hypothesis that evolutionary search exploits deterministic threshold discontinuities — related
• Sequencing rather than concept selection explains most of the EVOL–PPO-vanilla EP gap on ASSIST15 L=10 — related