I wouldn’t use torque for that unless you absolutely must have physics simulation - torque is notoriously fiddly to work with. I recently wrote a PID controller and while it kind of works fine, it’s still a wee bit off.
Why not rInterp to target rotation? It will be buttery smooth, give you plenty of fine granular control over the curve and you can even sample an external curve if you’re not happy with the built-in Ease functions.