Google DeepMind @_philschmid 分享:删掉 90% 的 Harness 后效果反而更好
Google DeepMind 的 @_philschmid 在 AI Engineer Code Summit 2026 上讲了个反直觉的做法:把评估 harness 删掉九成,钱省了,结果还差不多。正在搭 LLM 评估流程的人可以对照录像检查自己的管线里有多少冗余环节。
在 AI Engineer Code Summit 2026 上,Google DeepMind 团队的 @_philschmid 发表演讲「I deleted 90% of my harness (And It Got Better)」,提出把评估 harness 中的 90% 代码删掉后,成本下降且结果影响不大。
演讲的核心观点是:多数团队在搭建 AI 评估流程时引入了大量不必要的抽象层和检查逻辑,这些代码不仅拖慢迭代速度,还会掩盖真实信号。@aiDotEngineer 官方已将完整录像和文字稿整理发布,开发者可以对照自己的评估管线,排查哪些环节可以裁剪,用更精简的 harness 获得同样可靠的评估结果。
在刚刚结束的 AI Engineer Code Summit 2026 中,来自 Google Deepmind 团队的 @_philschmid 带来了分享「I deleted 90% of my harness (And It Got Better)」 Harness 删掉 90%,钱省下来了,结果影响也不大? @aiDotEngineer 官方把录像和文字稿都整理好了,赞,学起来! x.com/i/article/2122…📊 7 ⚡
This report is categorized under AITOP's "How-to" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Harness, AI Engineer Code Summit, philschmid, LLM evaluation.
Source attribution: curated from shao__meng (@shao__meng).