Is There a Transfer Benefit to Learning the Same Task?

I had this strong intuition that you should see some kind of transfer benefit to learning the same task on both sides. And it's a kind of tricky thing. I mean, what is it for tasks to be similar when they're non liner is kind of complicated. The general finding i had was that basically, if you train one side and then you try to train the other side to do the same task, you just get increasing transfer  in some sense. It's not very interesting because, so i haven't done this experiment, but i think that if i set up a task that had a different structure, bith the same signal strength, by that metric, it would get the