TY  - RPRT
TI  - A Single-Loop, Constant-Batch First-Order Penalty Method for Stochastic Bilevel Optimization
AU  - Xingyu Chen
AU  - Ming Yang
AU  - Quanqi Hu
AU  - Tianbao Yang
PY  - 2026
UR  - https://arxiv.org/abs/2610.07290
ID  - 2610.07290
ER  -