An introduction to training agents by updating model weights, covering supervised fine-tuning as a cold-start stage before reinforcement learning, what trajectories teach models, and why SFT remains necessary.

Save to open this file