I believe Add Scalar takes two inputs and simply adds them together. A Constant Scalar is just a number. So if I plugged two Constant Scalars set at 3 and 9 into the inputs of an Add Scalar, the Add Scalar would output 12.
Complement Scalar performs the function 1-x, where x is the input. So if you plugged 0.2 into a Complement Scalar, it would output 0.8. If you input -0.1, it would output 1.1 etc...
The Gain Scalar is slightly more complicated. Look at the graph for the Gain Scalar near the beginning of the thread. The X axis of the graph represents the input (what is plugged into the Gain Scalar) and the Y axis represents the output (what comes out of the Gain Scalar). If the Gain Scalar value is set at 0.8, then low input values close to 0 are made even lower and high input values close to 1 are made even higher but any input of 0.5 stays at 0.5.
If the Gain Scalar is set at 0.2, it's kind of the opposite - inputs close to 0 are pushed higher and inputs close to 1 are pushed lower. Essentially, just read off the graph. The same is true for the Bias Scalar.