Just more training.
Same training data used as in V1, just more training.
Untested, use with caution!