https://judyailab.com/tags/%E9%81%9E%E8%BF%B4%E8%87%AA%E6%88%91%E6%94%B9%E9%80%B2/ Recent content in 遞迴自我改進 on Judy AI Lab https://judyailab.com/logo.jpg https://judyailab.com/logo.jpg Hugo -- 0.147.4 zh-tw Sun, 20 Sep 2026 06:01:19 +0000 https://judyailab.com/posts/google-dream-rsi-agents-dream-self-improve/ Thu, 17 Sep 2026 11:00:00 +0000 https://judyailab.com/posts/google-dream-rsi-agents-dream-self-improve/ Google發布Dream-RSI,讓AI Agent做完任務後「做夢」複盤過去成敗、推演更好策略用到下一輪,免重跑任務即可遞迴自我改進。Gemini 3.1 Pro實測,同一演算法任務調用次數從550降到317,結果還更快。本文以實際帶6個AI Agent團隊的經驗,拆解進步關鍵是方法複利,而非換更大的模型。