It's not whether the signal is coded as voltage or as current what makes it more or less robust, it's the power involved.
Receiver of voltage signals (e.g. CMOS gate input) have extremely high impedance. So a noise voltage source does not need to provide much power to disturb the signal. That's why you shouldn't use it for "long" (some 0.1m) distances, because changing magnetic fields through the area circumscribed by the signal and GND line can easily induce some volts; although not much power (BTW: that's why twisted pairs are used; it causes the noise voltages of each twist to roughly cancel out each other).
Current signals (e.g. 4-20mA) are working against some resistors at the receiver side that require quite a voltage to generate the current. That means noise (voltage or current) source need to provide considerable power to have an effect. This is what makes it more robust.
But as thobie mentioned there are other good ways of reducing noise interference than just by requiring more power:
E.g. using differential signaling. That way you can still use voltage signals with high impedance receivers (but need as second signal line).
That's what is used e.g. for USB.