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🧵
