Q-Shaped Options Improve State and Action Abstraction in HRL

New Q-Shaped Options address the trade-off in Hierarchical Reinforcement Learning between maintaining necessary action distinctions and achieving coarse state abstraction. The method aims to reduce decision horizons while enabling better data aggregation across varying temporal scales.