Like averaging, weighted averaging is also used for regression tasks. Alternatively, it can be used while estimating probabilities in classification problems. Base learners are assigned different weights, which represent the importance of each model in the prediction.
A weight-averaged model should always be at least as good as your best model.