Gates that can go backward

Because they never change the number of qubits in the register

Because they are represented by unitary transformations

Because every gate is its own inverse when applied twice

Because measurement can undo any gate that came before it