Gimitest: A Comprehensive Tool for Testing Reinforcement Learning Policies

2026-07-08

AI Summary

This work addresses the limited scope of existing methods for evaluating the safety and robustness of reinforcement learning policies, which often suffer from insufficient coverage across scenarios, environments, and algorithms. To overcome these limitations, the authors propose Gimitest, a general-purpose, modular testing framework that enables unified, systematic evaluation of both single-agent and multi-agent policies across diverse environments and algorithms. Implemented in Python, Gimitest is compatible with mainstream libraries such as Gymnasium and PettingZoo, and supports dynamic modification of environment components alongside customizable test scenarios. Empirical results demonstrate that Gimitest effectively uncovers policy vulnerabilities under various perturbations and adversarial conditions, substantially enhancing the generality, flexibility, and scalability of reinforcement learning policy testing.

Abstract

Reinforcement learning (RL) policies can be unsafe and vulnerable to attacks. Ensuring their reliability is often a pain point as existing automated testing methods target only selected environments, testing scenarios, and RL algorithms. To address this, we propose a comprehensive framework for testing single- and multi-agent RL policies under varying conditions. Our implementation of this framework, Gimitest, is an open-source tool that supports various gym frameworks and allows for modifications of their integrated components. This article describes the framework and details Gimitest's functionality and architecture. It showcases its effectiveness in testing multiple RL policies in environments such as the official Farama Gymnasium and PettingZoo.

Problem

• Reinforcement Learning
• Policy Testing
• Safety
• Vulnerability
• Automated Testing

Innovation

• reinforcement learning testing
• multi-agent systems
• Gymnasium compatibility
• open-source framework
• policy robustness