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

TL;DR

SWE-MiniSandbox introduces a lightweight, container-free approach for scalable reinforcement learning in software engineering, significantly reducing system overhead while maintaining performance.

Contribution

It proposes a novel container-free method using kernel-level isolation and environment pre-caching to improve scalability and efficiency in RL training for SWE agents.

Findings

Disk usage reduced to 5% of container-based pipelines

Environment setup time decreased to 25% of baseline

Evaluation performance comparable to container-based methods

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.