TY  - RPRT
TI  - Learning Weakly Communicating Average-Reward CMDPs: Strong Duality and Improved Regret
AU  - Kihyun Yu
AU  - Beomhan Baek
AU  - Dabeen Lee
PY  - 2026
UR  - https://arxiv.org/abs/2605.11586
ID  - 2605.11586
ER  -