Mage-VL-Cowork：从演化 harness 学习混合计算机使用的完整训练配方
原标题：Mage-VL-Cowork: A Full Training Recipe to Learn Hybrid Computer Use From Evolved Harnesses
AI 摘要
该论文提出 Mage-VL-Cowork，一个 4B 参数混合计算机使用智能体（CUA），从通用多模态对齐检查点出发，经五阶段渐进式课程（中训练、SFT、RL）训练，同时结合 GUI 视觉交互与终端代码执行。为支撑大规模训练，作者构建 Mage-CUA-Forge 环境合成流水线，覆盖 6,915 项技能、合成 293,806 个桌面与网页任务，并将单任务重复开销降低 82.6%。团队还演化教师 harness 以生成高质量轨迹，最终模型在 OSWorld-Verified 上达到 66.0%，在同参数规模下取得最优表现。
正文节选
October, 2026 Contents Abstract Effective computer-use agents (CUAs) should seamlessly combine visual GUI interaction with code and terminal execution within a single policy. We present Mage-VL-Cowork, a 4B hybrid CUA trained from a general multimodally aligned checkpoint through an end-to-end, progressive five-stage curriculum across midtraining, SFT, and RL. To power this curriculum at scale, Mage-CUA-Forge replaces per-task agent construction with reusable, skill-conditioned generation workfl