The creation of innovative learning environments in courses to provide sustained talent support has long remained a central research concern for high-quality social development. Neglected foundational skills in professional course clusters are often a hidden barrier to higher education for sustainable development (HESD). To close five persistent foundational skill gaps across improper citation (J1), ineffective figure use (J2), poor analysis (J3), irresponsible AI use (J4), and comprehensive application (J5) within the hydraulic engineering course cluster, a four-stage deliberate practice module (5Di-40Pr-5Tr-3Cm) has been embedded into a two-week hydraulic model experiment course, and its learning outcomes are systematically evaluated. A systematic analysis of its achievement levels across neglected foundational skill indicators of J1~J5 at each stage was conducted, stratified by the overall cohort and subgroups (P: objective demand, T: behavior type, G: optimization methods). The key findings include: ① deliberate practice demonstrates better teaching outcomes than lecture-based instruction, which can be evidenced in 2026, when J5’s achievement levels at the 3Cm stage yielded a moderate effect size relative to the 2025 lecture-based condition (d = 0.42); compared with the 2024 no-intervention baseline, the cumulative effect is a obvious increasing trend (d = 1.43); ② In far-transfer subgroup diagnosis, P2 (medium objective demand) shows a rank-order reversal, low at 40Pr but higher at 3Cm, and is identified as the “partial understanding” group and providing a diagnostic anchor for tiered intervention; ③ In near-transfer pathway diagnosis, J5’s low performance in 5Tr (65.35%, below overall mean of 83.09%; CV = 7%) stems from two distinct pathways: a “knowledge-deficit pathway” (max-decay subgroups) and a “processing-load pathway” (subgroups where J1, J2 do not exhibit max decay). In addition, stage-specific thresholds (40Pr: 90%, range 60~99%; 5Tr and 3Cm: 83% ± 3%, range of for 40Pr, mean = 90%, recommended range = 60~99%; for 5Tr, mean = 83% ± 3%, range = 65~96% and 75~90%) provide quantitative benchmarks for targeted intervention. These cumulative findings are intended to advance the evaluation paradigm of engineering practice courses from “total score attainment” toward “structural diagnosis” and align with the competency-oriented philosophy of higher education for sustainable development (HESD).

Reliable prediction of post-initiation run-out distance of rainfall-induced landslides is essential for hazard assessment, evacuation planning, and disaster-risk mitigation. However, most existing data-driven approaches formulate run-out prediction as a deterministic regression problem and therefore provide limited information on predictive uncertainty, rare long-runout events, and explainable decision support. To address these limitations, this study proposes CQR-EVT-XAI, a trustworthy AI framework that integrates Quantile LightGBM, Conformalized Quantile Regression (CQR), Extreme Value Theory (EVT), and Explainable Artificial Intelligence (XAI) for uncertainty-aware and explainable landslide run-out risk prediction. Based on 10,158 rainfall-induced landslide samples, physics-informed features are constructed from elevation difference H, source area A, source volume V, and mean slope angle $\theta$. The proposed framework generates calibrated prediction intervals, threshold-based exceedance probabilities, upper-tail risk indicators, and interpretable risk levels. The CQR-LightGBM median model achieves high point-prediction accuracy, with $R^2$ = 0.939, RMSE = 18.03 m, and MAE = 6.55 m. Conformal calibration improves the empirical coverage of the nominal 90% and 95% prediction intervals from 0.813 to 0.903 and from 0.876 to 0.953, respectively. Tail-risk analysis shows that the upper prediction bound ${\widehat{L}}_{95}$ effectively identifies extreme long-runout events, achieving recall values of 0.974 and 0.900 for L > 300 m and L > 500 m, respectively. SHAP analysis reveals that elevation difference H, source volume V, and energy-related derived features dominate both median run-out prediction and upper-tail risk behavior, while slope-related variables mainly influence predictive uncertainty and exceedance-risk levels. These results demonstrate that the proposed CQR-EVT-XAI framework provides a practical workflow for calibrated uncertainty quantification, tail-risk identification, and explainable decision support in rainfall-induced landslide run-out risk assessment.