微软开源 Agent Lightning v1.0：3,500 行轻量级 Agentic RL 框架
原标题：Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses
AI 摘要
微软亚洲研究院开源了 Agent Lightning v1.0，一个约 3,500 行代码的轻量级 Agentic RL 框架，提出 Harnessed Agentic RL 训练范式，让部署时使用的 agent harness 直接参与强化学习，无需在训练框架内重新实现 agent。框架原生支持 Kubernetes，并给出端到端编码 agent 训练示例：基于约 6,000 条开源数据样本，将 Qwen3.5-9B 在 SWE-bench Verified 上的 Pass@1 从 41.8% 提升至 56.4%。
正文节选
At a glance - Harnessed Agentic RL: Microsoft Research Asia introduces a training paradigm in which the same agent harness used in deployment participates directly in reinforcement learning, removing the need to reimplement the agent inside the training framework. - Lightweight by design: Agent Lightning v1.0 delivers a complete agent RL control plane in roughly 3,500 lines of code. - Native Kubernetes support: agents run as standard Kubernetes jobs on self-managed clusters, cloud Kubernetes, or