OmniSafe

OmniSafe is a benchmark that contains constrained reinforcement learning environments used to evaluate safety in reinforcement learning algorithms.

Papers using OmniSafe (3)

parallelcbf: A composable safety-filter and auditability framework for tensor-parallel reinforcement learning2026Balance Reward and Safety Optimization for Safe Reinforcement Learning: A Perspective of Gradient Manipulation2024 · 9 citesEnhancing Efficiency of Safe Reinforcement Learning via Sample Manipulation2024