Comment by julia-zheng

For the most part the model should generalize pretty well to new formats, but we do constantly monitor and update to catch up with any holes.