Primary sourceelvis
View sourceEvent ·Single source
MemoHarness：优化agent控制层的六维可编辑框架
论文提出 MemoHarness，将 LLM agent 的 harness（外部控制层）分解为上下文、工具、生成、编排、记忆、输出六个可编辑面。它通过检索相似案例自动为每个新场景调整 harness，无需测试时标签或梯度更新。在 shell-agent 基准测试中，MemoHarness 达到 0.806 分，优于最强固定 harness 基线的 0.722。每任务成本低于最强商业基线。
Current read
想自己调agent的提示和流程？这篇论文把控制层拆成六块，不用额外标签就能自动优化，比固定harness效果好不少。
2 sources53° AI heatLast updated 2026/7/17 14:20:02
Evidence
2 sourcesRelated source 1shao__meng
View sourceConflict 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.