std::tuple of two type VS. std::pair [duplicate]

This question already has an answer here:

• Difference between std::pair and std::tuple with only two members? 4 answers

Is there any advantage of using std::pair over std::tuple (or vice versa) when I need only two different types?

For example:

std::pair<int,double> foo;
std::tuple<int,double> bar;

I can name one advantage is that if I need, in future, to add a third type, std::tuple can serve me better than pair because I have to change the std::pair to std::tuple or to work around it (like making a struct that contains two of them).

By advantages, I mean performance, readability, usability, maintainability or anything...