Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control

Researchers introduced Reflex, a reinforcement learning framework that leverages reflection symmetry for state-based continuous control.

Reflex applies axial and bilateral reflections to improve policy learning in environments such as OpenAI Gym and DeepMind Control.

When combined with PPO and SAC, the method shows better sample efficiency and stronger benchmark performance.

The work extends symmetry-based RL beyond image and rotation settings to reflection in state-driven control tasks.