How does the random$-$restart hill climbing algorithm improve upon the basic hill climbing approach?

Select one:

By restarting the algorithm from different random states until a goal state is found.

By only choosing the best possible move at each step.

By ensuring each step moves directly toward the goal.

By reducing the search space systematically.

By using backtracking to avoid local optima.