A practical map of LLM post-training: how SFT, reward models, RL (PPO, GRPO), DPO, and RLVR fit together, and why a reward model is not RL.