SafeRLEval：安全强化学习的新评测指标体系
Evaluation Metrics for Safe Reinforcement Learning
AI 导读
研究者针对安全强化学习（RL）提出一套新评测指标，指出现有基准只报告平均安全性，无法刻画安全约束（CMDP）被违反的频率与严重程度。新框架包括跨任务与安全界限聚合的评测指标、用于分类比较算法的安全等级体系，以及在多个安全导航任务上的实证评估。研究建议联合报告聚合指标、分布信息和任务特定结果，并开源了评测套件 SafeRLEval。
正文 · 原文
Abstract: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.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2609.15315 [cs.AI] |
| (or arXiv:2609.15315v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.15315 arXiv-issued DOI via DataCite |
Submission history
From: Lindsay Spoor [view email]
[v1]
Mon, 14 Sep 2026 10:04:22 UTC (1,381 KB)
[v2]
Thu, 29 Oct 2026 15:11:46 UTC (1,406 KB)
来源：arXiv cs.AI · arxiv.org