SWE-MiniSandbox: Container-Free Reinforcement Learning for Building Software Engineering Agents

2026-02-11

AI Summary

This work proposes a lightweight, container-free training framework for reinforcement learning–based software engineering agents, addressing the high storage overhead, slow environment initialization, and complex permission management inherent in conventional task-level containerization. By leveraging kernel-level isolation, lightweight workspace management, and environment pre-caching techniques, the framework enables efficient and scalable agent training without relying on containers. Experimental results demonstrate that the approach reduces disk usage to approximately 5% of that required by standard container-based solutions and accelerates environment setup time to about 25% of the original duration, while maintaining comparable performance. These improvements substantially enhance resource efficiency and training scalability, offering a practical alternative for large-scale reinforcement learning applications in software engineering.

Technology Category

• Machine Learning: Reinforcement Learning
• Multiagent Systems: Adversarial Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Reinforcement learning (RL) has become a key paradigm for training software engineering (SWE) agents, but existing pipelines typically rely on per-task containers for isolation. At scale, pre-built container images incur substantial storage overhead, slow environment setup, and require container-management privileges. We propose SWE-MiniSandbox, a lightweight, container-free method that enables scalable RL training of SWE agents without sacrificing isolation. Instead of relying on per-instance containers, SWE-MiniSandbox executes each task in an isolated workspace backed by kernel-level mechanisms, substantially reducing system overhead. It leverages lightweight environment pre-caching techniques to eliminate the need for bulky container images. As a result, our approach lowers disk usage to approximately 5\% of that required by container-based pipelines and reduces environment preparation time to about 25\% of the container baseline. Empirical results demonstrate that SWE-MiniSandbox achieves evaluation performance comparable to standard container-based pipelines. By removing the dependency on heavy container infrastructure, SWE-MiniSandbox offers a practical and accessible foundation for scaling RL-based SWE agents, particularly in resource-constrained research environments.

Problem

• reinforcement learning
• software engineering agents
• container overhead
• environment isolation
• scalable training

Innovation

• container-free
• reinforcement learning
• software engineering agents
• lightweight sandboxing
• kernel-level isolation