The field of control and reinforcement learning is moving towards the development of more efficient, interpretable, and expressive models. Researchers are exploring alternative architectures, such as discrete logic circuits and generative policies, to improve performance and stability in complex control tasks. The use of novel control strategies, such as density-based predictive control, is also being investigated for efficient non-uniform area coverage. Noteworthy papers in this area include: Differentiable Weightless Controllers, which introduces a symbolic-differentiable architecture for continuous control, and GoRL, which presents a framework for online reinforcement learning with generative policies that outperforms existing baselines.