A key finding was that most AutoML tools tended to favor tree-based models and ensembles, which often delivered high accuracy but raised concerns about interpretability and overfitting. The study ...