AI RESEARCH
Path-Coupled Bellman Flows for Distributional Reinforcement Learning
arXiv CS.AI
•
ArXi:2605.08253v1 Announce Type: cross Distributional reinforcement learning (DRL) models the full return distribution, but existing finite- or quantile-based methods rely on projections, while recent flow-based approaches can suffer from \emph{boundary mismatch} at the flow source or from \emph{high-variance} bootstrapping when current and successor noises are independent.