cs.LGMay 7, 2026

Path-Coupled Bellman Flows for Distributional Reinforcement Learning

Authors: Boyang Xu, Qing Zou, Siqin Yang, Hao Yan

Organizations: School of Computing and Augmented Intelligence, Arizona State University, Tempe, AZ, USA.

Abstract

Distributional reinforcement learning (DRL) models the full return distribution, but existing finite-support 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. We propose Path-Coupled Bellman Flows (PCBF), a continuous-time DRL method that learns return distributions with flow matching using \textbf{source-consistent Bellman-coupled paths}: the current path starts from the required base prior at t=0t{=}0, reaches the Bellman target at t=1t{=}1, and maintains a pathwise affine relation to the successor flow at intermediate times (without requiring time-tt marginals to satisfy a distributional Bellman fixed point for all tt). PCBF couples current and successor return flows through shared base noise and uses a λλ-parameterized control-variate target: λ=0λ{=}0 recovers an unbiased sample Bellman target, while λ>0λ{>}0 trades controlled bias for variance reduction. Experiments on analytically tractable MRPs, OGBench, and D4RL show improved distributional fidelity and training stability, and competitive offline RL performance.

Explore similar work

CardsList
  1. Quantile-Coupled Flow Matching for Distributional Reinforcement Learning

    May 8, 2026Michael Groom, Victor-Alexandru Darvariu, Lars Kunze +2Bellman Equation

  2. On Distributional Reinforcement Learning in Chaotic Dynamical Systems

    May 28, 2026James Rudd-Jones, Mirco Musolesi, María Pérez-OrtizChaosBellman Equation