Hacker News new | ask | show | jobs
by ianand 453 days ago
The model architecture is the same during RL but the training algorithm is substantially different.