reinforcement learning
Reinforcement Learning’s Real Bottlenecks: Environments and Rewards, Not Algorithms
RL’s key bottlenecks are shifting from algorithm choice to environment design, reward systems, and sample efficiency—not necessarily better algorithms