Guardian: Decoupling Exploration from Safety in Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cai, Kaitong, Zhang, Jusheng, Yang, Jing, Wang, Keze
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917044660731904
author Cai, Kaitong
Zhang, Jusheng
Yang, Jing
Wang, Keze
author_facet Cai, Kaitong
Zhang, Jusheng
Yang, Jing
Wang, Keze
contents Hybrid offline--online reinforcement learning (O2O RL) promises both sample efficiency and robust exploration, but suffers from instability due to distribution shift between offline and online data. We introduce RLPD-GX, a framework that decouples policy optimization from safety enforcement: a reward-seeking learner explores freely, while a projection-based guardian guarantees rule-consistent execution and safe value backups. This design preserves the exploratory value of online interactions without collapsing to conservative policies. To further stabilize training, we propose dynamic curricula that gradually extend temporal horizons and anneal offline--online data mixing. We prove convergence via a contraction property of the guarded Bellman operator, and empirically show state-of-the-art performance on Atari-100k, achieving a normalized mean score of 3.02 (+45\% over prior hybrid methods) with stronger safety and stability. Beyond Atari, ablations demonstrate consistent gains across safety-critical and long-horizon tasks, underscoring the generality of our design. Extensive and comprehensive results highlight decoupled safety enforcement as a simple yet principled route to robust O2O RL, suggesting a broader paradigm for reconciling exploration and safety in reinforcement learning.
format Preprint
id arxiv_https___arxiv_org_abs_2510_22859
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Guardian: Decoupling Exploration from Safety in Reinforcement Learning
Cai, Kaitong
Zhang, Jusheng
Yang, Jing
Wang, Keze
Machine Learning
Artificial Intelligence
Hybrid offline--online reinforcement learning (O2O RL) promises both sample efficiency and robust exploration, but suffers from instability due to distribution shift between offline and online data. We introduce RLPD-GX, a framework that decouples policy optimization from safety enforcement: a reward-seeking learner explores freely, while a projection-based guardian guarantees rule-consistent execution and safe value backups. This design preserves the exploratory value of online interactions without collapsing to conservative policies. To further stabilize training, we propose dynamic curricula that gradually extend temporal horizons and anneal offline--online data mixing. We prove convergence via a contraction property of the guarded Bellman operator, and empirically show state-of-the-art performance on Atari-100k, achieving a normalized mean score of 3.02 (+45\% over prior hybrid methods) with stronger safety and stability. Beyond Atari, ablations demonstrate consistent gains across safety-critical and long-horizon tasks, underscoring the generality of our design. Extensive and comprehensive results highlight decoupled safety enforcement as a simple yet principled route to robust O2O RL, suggesting a broader paradigm for reconciling exploration and safety in reinforcement learning.
title Guardian: Decoupling Exploration from Safety in Reinforcement Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.22859