Event ·Multi-source

Google Research 发布 Harness²:让智能体在冻结模型外自行改进运行框架

Google Research 发布了 Harness² 框架,在模型参数冻结的情况下迭代改进围绕它的 harness(运行环境)。流程分四步:先用基础 harness 跑任务记录基线轨迹,Editor agent 起草 2 个候选编辑(SI、Skills、hooks 等)并运行,Composer agent 对比新轨迹与基线后合并有效编辑,Reflector agent 把经验存入库供后续任务调用。Gemini 3.5 Flash 在 WorkBuddy-Bench 上提升 8.4 分,JobBench 上提升 7.5 分,且在全部 3 个基准上都因自我编辑 harness 而提升。消融显示 Skills 和脚本贡献约 50% 的编辑,去掉 Skills 会损失 5.3 分。论文与代码已公开。

Current read

Google 把 harness 本身交给 agent 改:Editor 出编辑、Composer 合并、Reflector 存经验,Gemini 3.5 Flash 三个基准全涨,做 agent 工程的可以看看代码。

5 sources47° AI heatLast updated 2026/10/19 11:58:08

Evidence

5 sources
Primary sourcePhilipp Schmid
View source
Related source 1marktechpost
View source
Related source 2shao__meng
View source
Related source 3Shubham Saboo
View source
Related source 4Google Research
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.