Mode-Dependent Rectification for Stable PPO Training

Abstract:Mode-dependent architectural components (layers that behave differently during training and evaluation, such as Batch Normalization or dropout) are commonly used in visual reinforcement learning but can destabilize on-policy optimization. We show that in Proximal Policy Optimization (PPO), discrepancies between training and evaluation behavior induced by Batch Normalization lead to policy mismatch, distributional drift, and reward collapse. We propose Mode-Dependent Rectification (MDR), a lightweight dual-phase training procedure that stabilizes PPO under mode-dependent layers without architectural changes. Experiments across procedurally generated games and real-world patch-localization tasks demonstrate that MDR consistently improves stability and performance, and extends naturally to other mode-dependent layers.

Machine Learning (cs.LG); Artificial Intelligence (cs.AI)

arXiv:2602.05619 [cs.LG]

(or arXiv:2602.05619v2 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2602.05619 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.