Reproducibility study on how to find Spurious Correlations, Shortcut Learning, Clever Hans or Group-Distributional non-robustness and how to fix them

Reproducibility study unifying fragmented research on spurious correlations in DNNs across frameworks like distributionally robust optimization, invariant risk minimization, and explainable AI. Compares correction methods under real-world constraints and finds XAI-based approaches generally outperform non-XAI baselines for ensuring models rely on causally relevant features.