AI RESEARCH
Proximal Policy Optimization in Path Space: A Schr\"odinger Bridge Perspective
arXiv CS.LG
•
ArXi:2603.21621v1 Announce Type: new On-policy reinforcement learning with generative policies is promising but remains underexplored. A central challenge is that proximal policy optimization (PPO) is traditionally formulated in terms of action-space probability ratios, whereas diffusion- and flow-based policies are naturally represented as trajectory-level generative processes. In this work, we propose GSB-PPO, a path-space formulation of generative PPO inspired by the Generalized Schr\"odinger Bridge.