pre-training, post-training, fine-tuning, RL, synthetic data generation, human intelligence, evaluation