TY  - RPRT
TI  - Off-Policy Reinforcement Learning with Anytime Safety Guarantees via Robust Safe Gradient Flow
AU  - Pol Mestres
AU  - Arnau Marzabal
AU  - Jorge Cortés
PY  - 2025
UR  - https://arxiv.org/abs/2510.01492
ID  - 2510.01492
ER  -