同一模型不同 Harness：CUGA、Hermes 与 Deep Agents 对比
原标题：Same Model, Different Harness: CUGA, Hermes, and Deep Agents — When to Pick CUGA?
AI 摘要
IBM 开源了 CUGA（ConfigUrable Generalist Agent）harness，并在 AppWorld 基准的 90 个任务上，用同一开源模型 GLM-5.3-flash 对比了 CUGA、Nous Research 的 Hermes 和 LangChain 的 Deep Agents。结果显示，仅更换 harness 就让任务完成率从 30% 提升到 81%，三个 harness 全部通过的场景数分别为 20、8、2。CUGA 面向企业场景，将支付审批、PII 处理等业务规则从提示词移到 harness 层强制执行，支持在自有环境和隔离部署中运行。
正文节选
CUGA Policy Playground Two live CUGA agents, with and without policies There are nearly 200 agent harnesses to choose from. We gave three of them the same open-weight model and 90 AppWorld tasks, and changing the harness moved task completion from 30% to 81%. For enterprise use, CUGA adds what this benchmark does not test: rules that the harness enforces and deployment inside your own boundary. Community lists such as awesome-agent-harness and best-of-Agent-Harnesses track nearly 200 of them, an