Ensembled methods
They average out biases, reduce variance, and are less likely to overfit.
There’s a common line in machine learning which is: “ensemble and get 2%.”
This implies that you can build your models as usual and typically expect a small performance boost from ensembling.