Comment by trifurcate

Why are you ruling out the possibility that training on the material may confer an advantage when the data is presented, even if the advantage may not be strong enough to pass the test without the data present in the context window?