CodeMidas：从代码本身扩展智能体编码强化学习环境
原标题：Paper page - CodeMidas: Scaling Agentic Coding RL Environments from Code Itself
AI 摘要
研究团队提出 CodeMidas，一个以源代码为唯一任务特定输入的智能体流水线，将现有代码库中已实现的功能转化为可执行的强化学习环境。该方法通过智能体探索功能、构建基于原始代码执行的测试，并经过执行检查与多次解答采样来验证和筛选任务。最终数据集包含来自 3185 个开源代码库、覆盖 23 种编程语言和 15 个技术领域的 5545 个训练任务。用 GRPO 训练 MiMo-V2.5 后，在五个基准上均有提升，包括 DeepSWE +11.7%、ProgramBench +17%、Terminal-Bench v2.1 +8.5%。
正文节选
CodeMidas: Scaling Agentic Coding RL Environments from Code Itself Abstract Training capable coding agents via reinforcement learning (RL) requires diverse tasks with reliable verifiers. Open-source codebases offer a rich source of such tasks, while existing methods typically rely on development artifacts such as issues and commits, limiting the range of tasks that can be extracted. To better scale RL environments, we present CodeMidas, an agentic pipeline that turns implemented functionality in