Controlling the LEDs is as simple as bringing on pin high, and another pin low, sometimes in software this is represented in 1 (high), and 0(low). When you multiplex you have rows and columns. Lets say you have a grid of 8 by 8 (64 LEDs). All your cathodes connect in row are connected, and all the anodes in a column are connected, this leaves you with 8 anodes, and 8 cathodes to deal with (instead of 64 anodes and cathodes to deal with).
When multiplexing you only light 1 led at a time, but you can turn many leds on and off very quickly, so that it appears that they are all lit at the same time, this is called persistence of vision.
So, to light a single LED, you turn one anode pin high, and one cathode pin low, this lights one LED, then you pause for a few milliseconds, then turn those pins off,
You repeat this process until you light up whatever pattern you need.
This is how every cube or matrix works.
A 2x3 matrix would require 5 pins from your controller, multiplexing is only saving you 1 pin over controlling the anodes or cathodes independently, so its realistically only practical as an learning example.