As AI agents face increasingly long and complex tasks, decomposing them into subtasks becomes increasingly appealing.

But how do we discover such temporal structure?

Hierarchical RL provides a natural formalism-yet many questions remain open.

Here's our overview of the field🧵