topic

reinforcement-learning

PPO Entropy Collapse

Why clipped policy updates can still lose exploration pressure, and what to watch during training.