How do you smooth/lerp "Get Input Motion State" outputs over time?

You could try adding a springarm and enabling lag:

214349-lag.png

Another solution is to interpolate manually, by updating the target and letting Tick do the rest: