“Temporal Difference (TD) learning enables an agent to handle sparse, long-term rewards by using a value function to predict the final outcome and reinforcing intermediate actions that improve that prediction.”