Evaluation Metrics for Safe Reinforcement Learning
On the map: Deep reinforcement learning methods → Safe and constrained RL
Abstract (arXiv)
Safe reinforcement learning (RL) is commonly formalized as a Constrained Markov Decision Process (CMDP), in which an agent maximizes expected reward while keeping its expected cumulative cost below a specified safety bound. Existing safe RL benchmarks predominantly report whether an algorithm is safe on average, following this expectation-based guarantee. We argue that this convention is insufficient to reliably characterize an algorithm's true safety: it fails to capture how often and how severely the safety bound is violated, whether this holds consistently across tasks and safety bounds, and whether training-time behavior is representative of behavior of the final converged policy. Therefore, we introduce (i) evaluation metrics for safe RL that address each of these concerns and in addition allow for aggregation across tasks and safety bounds. We furthermore define (ii) a safety tier system to systematically categorize and compare algorithms in terms of safety and reliability at both training and for a final policy. Using this framework, we provide (iii) an empirical safety evaluation across multiple safety navigation tasks. Our results show that aggregate metrics, distributional reporting, and task- and safety bound-specific results each reveal information the other metrics cannot. We therefore recommend reporting all three jointly, rather than compressing this information into a single value, as is common practice. We provide SafeRLEval, an open-source evaluation suite to support the reliable characterization of safety in future safe RL research.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safe and constrained RL, one of 15 directions in Deep reinforcement learning methods. Builds on 13 of its 24 references that are in the map. First posted on arXiv 14 Sep 2026.
Builds on
13 of 24 references.
Closest ancestors · 2
Same direction, newest first
- 2026CRAX: Fast Safe Reinforcement Learning Benchmarking18 Jun
- 2024ActSafe: Active Exploration with Safety Constraints for Reinforcement Learning12 Oct
Other references · 11
- 2017Proximal Policy Optimization Algorithmscited by 15,215
- 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,885
- 2021Deep Reinforcement Learning at the Edge of the Statistical Precipicecited by 807
- 2017Safe Reinforcement Learning via Shieldingcited by 449
- 2018Reward Constrained Policy Optimizationcited by 417
- 2020Responsive Safety in Reinforcement Learning by PID Lagrangian Methodscited by 316
- 2019End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Taskscited by 289
- 2018Safe Exploration in Continuous Action Spacescited by 247
- 2023Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmarkcited by 143
- 2023OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Researchcited by 80
- 2025HASARD: A Benchmark for Vision-Based Safe Reinforcement Learning in Embodied Agentscited by 6
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2024Revisiting Safe Exploration in Safe Reinforcement learning
- 2025Provably Optimal Reinforcement Learning under Safety Filtering
- 2024Safe Reinforcement Learning with Learned Non-Markovian Safety Constraints
- 2025A Survey of Safe Reinforcement Learning and Constrained MDPs: A Technical Survey on Single-Agent and Multi-Agent Safety
- 2024A Survey of Constraint Formulations in Safe Reinforcement Learning
- 2023Datasets and Benchmarks for Offline Safe Reinforcement Learning
- 2025Safety Modulation: Enhancing Safety in Reinforcement Learning through Cost-Modulated Rewards
- 2024Balance Reward and Safety Optimization for Safe Reinforcement Learning: A Perspective of Gradient Manipulation