MARPO: A Reflective Policy Optimization for Multi-Agent Reinforcement Learning
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (35) · 29740-29748
DOI: 10.1609/aaai.v40i35.40219 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Cuiling Wu und 3 weiteren Autor:innen.
Abstract (Originalsprache)
We propose Multi-Agent Reflective Policy Optimization MARPO to alleviate the issue of sample inefficiency in multi-agent reinforcement learning. MARPO consists of two key components: a reflection mechanism that leverages subsequent trajectories to enhance sample efficiency, and an asymmetric clipping mechanism that is derived from the KL divergence and dynamically adjusts the clipping range to improve training stability. We evaluate MARPO in classic multi-agent environments, where it consistently outperforms other methods.
Zitieren
APA
Wu, C., Gan, Y., Xing, J., & Fu, Y. (2026). MARPO: A Reflective Policy Optimization for Multi-Agent Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 40(35), 29740-29748. https://doi.org/10.1609/aaai.v40i35.40219
Vancouver
Wu C, Gan Y, Xing J, Fu Y. MARPO: A Reflective Policy Optimization for Multi-Agent Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(35):29740-29748. doi:10.1609/aaai.v40i35.40219
BibTeX
@article{Wu2026,
author = {Wu, Cuiling and Gan, Yaozhong and Xing, Junliang and Fu, Ying},
title = {MARPO: A Reflective Policy Optimization for Multi-Agent Reinforcement Learning},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {35},
pages = {29740--29748},
doi = {10.1609/aaai.v40i35.40219}
}
EUVIMED