3 ms·The model architecture is the same during RL but the training algorithm is substantially different.by ianand 2y agoThe model architecture is the same during RL but the training algorithm is substantially different.