Adjusted $R^2$ always increases.

Adjusted $R^2$ decreases if the new predictor does not improve the model enough.

$R^2$ always increases or stays the same.

Adjusted $R^2$ is always greater than $R^2$.