This is really cool! I'm curious to see how this performs against other AutoML systems like autogluon, autokeras, and AutoML Tables. Do you have any model quality benchmarks on standard research tabular datasets (UCI, OpenML, PMLB) or Kaggle competitions?