ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation

ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation | ResearchPod