TY  - RPRT
TI  - Multi-Step Alignment as Markov Games: An Optimistic Online Gradient Descent Approach with Convergence Guarantees
AU  - Yongtao Wu
AU  - Luca Viano
AU  - Yihang Chen
AU  - Zhenyu Zhu
AU  - Kimon Antonakopoulos
AU  - Quanquan Gu
AU  - Volkan Cevher
PY  - 2025
UR  - https://arxiv.org/abs/2502.12678
ID  - 2502.12678
ER  -