Radio-frequency (RF) neuromodulation systems require careful balance between therapeutic efficacy and patientsafety, particularly regarding specific absorption rate (SAR)exposure limits. This paper presents a constrained reinforcementlearning approach for closed-loop beamforming that minimizesSAR while maintaining neuromodulation utility. Using primaldual optimization, our method learns policies that respect safetybudgets through adaptive Lagrange multipliers. Experimentalresults demonstrate that beamforming with learned … Continue reading Safety Budgets for RF Neuromodulation: Closed-Loop Power Minimization with Reinforcement Learning