This project compares four RL algorithms on CartPole-v1 by evaluating their average return and episode length. The study covers online methods (REINFORCE, DQN, A2C) and an offline method (Decision Transformer) trained on expert data.