Here is benchmark done by independent team of researchers https://openml.github.io/automlbenchmark/ https://openml.github.io/automlbenchmark/ I think most of o

Here is benchmark done by independent team of researchers https://openml.github.io/automlbenchmark/ https://openml.github.io/automlbenchmark/ I think most of overfitting is avoided with early stoppoing technique. The underfitting can be avoidwd with using large training time.