td3

This paper introduces spectral effective-rank entropy as a metric to measure and control critic complexity in actor-critic reinforcement learning, demonstrating its measurability and controllability in TD3 and PPO experiments.