Google 开源 Harness2:测试时递归改写智能体 harness 的框架
Google Open-Sources Harness2, a Recursive Agent Harnessing Framework
Google 开源了个叫 Harness2 的框架,不改模型参数、只改测试时的 harness 代码,就让 GLM-5 在 WorkBuddy-Bench 上涨了 11.5 分,做智能体的可以看看。
Google Research 发布开源框架 Harness2,采用 Apache-2.0 许可证,核心思路是在测试时递归重写智能体的 harness 代码,不需要更新模型权重。在 WorkBuddy-Bench 基准上,一次 harness 精炼迭代就让 GLM-5 提升 11.5 分,全程零权重更新。这意味着同一个模型不改参数,只改外围执行框架就能获得明显收益。
Google Open-Sources Harness2, a Recursive Agent Harnessing Framework
Google Research released Harness2, an Apache-2.0 framework that rewrites an agent's harness at test time. One refinement step lifted GLM-5 by 11.5 points on WorkBuddy-Bench with no weight updates. The post Google Open-Sources Harness2, a Recursive Agent Harnessing Framework appeared first on MarkTechPost .