讓 AI 助理自己修改提示詞和記憶,可能改善新任務,卻也可能讓原本會做的事變差,或把錯誤內容留在記憶裡。
- Google 的論文提出 GUARD,保留底層模型不變,讓 AI 助理調整自己的提示詞、技能、工具規則和記憶。
- 每次修改前,先把任務拆成檢查專案,再比對修改前後的表現,優先處理新出現的問題。在 43 項託管 SDK 任務中,GUARD 的正確率為 80.80%,比起始設定高 8.89 個百分點。
- 在程式記憶任務中,更新正確率為 90.1%,且沒有記憶汙染。
AI 伴隨問答
針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:
GUARD 會先檢查修改提示詞與記憶後造成的行為變化,並優先修復新出現的問題,可用來理解自我改進 AI 助理如何降低舊能力受損的風險。
Banger paper from Google on self-improving agent harnesses.
(bookmark it)
It addresses common issues in self-improving agent harnesses like local overfitting, catastrophic forgetting, and memory pollution.
They propose GUARD, which keeps the base model frozen and lets the agent edit its own prompts, skills, tool rules, and memory. An edit is accepted only when its gains outweigh the behaviors it breaks.
- On 43 hosted SDK tasks, GUARD reached 80.80% accuracy, 8.89 points over the starting harness (p = 0.008).
- It beat Meta-Harness by 6.28 points and used 54.5% fewer tokens per task.
- On procedural memory tasks, it led update correctness at 90.1% with zero memory pollution. Aggressive rewriters scored similarly and tripled pollution.
How does it know what broke? Each task is split into self-checks before execution. Every round grades trajectories against them, pins rules the agent already handles, and ranks newly broken behaviors first. At test time, one frozen harness runs in a single pass.
Chat with Paper: https://t.co/DYRRAbMJMd