Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control
The paper introduces Reflex, a new approach to reinforcement learning that utilizes reflection symmetry in state-based continuous control tasks. This method aims to improve sample efficiency by integrating reflection symmetry into policy learning. The authors demonstrate that Reflex outperforms standard baselines when evaluated on various benchmarks.
- ▪Reflex leverages group-invariant Markov Decision Processes to enhance sample efficiency in reinforcement learning.
- ▪The paper focuses on state-based continuous control tasks and introduces two types of reflection: axial and bilateral.
- ▪Reflex integrates with both on-policy and off-policy reinforcement learning algorithms, showing superior performance in tests.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.23415 (cs) [Submitted on 22 May 2026] Title:Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control Authors:Shuai Zhen, Yifan Zhang, Yuling Wang, Yanhua Yu View a PDF of the paper titled Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control, by Shuai Zhen and 3 other authors View PDF HTML (experimental) Abstract:Reinforcement learning has long struggled with poor sample efficiency. One promising approach to mitigate this problem is leveraging group-invariant Markov Decision Processes ($G$-invariant MDPs).
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
