Post-training

The stage of the large-model training pipeline that follows pre-training. Through techniques such as reinforcement learning, it further shapes the model's behavioral policy and improves performance on specific tasks.