智能體記憶蒸餾：以階層式教師記憶賦能小型LLM智能體

摘要

記憶系統已展現出提升智能體效能的潛力，但其在小語言模型上的應用潛力仍 largely 未獲充分探索，因為小模型難以自行產生足夠數量的成功軌跡。我們提出智能體記憶蒸餾（Agent Memory Distillation, AMD），這是一個免訓練框架，透過層級式記憶將結構化知識從大型教師智能體轉移至小型學生智能體。AMD 從成功的教師軌跡中建構三種互補的記憶類型：工作流程記憶編碼任務層級的策略，子任務記憶在中間粒度提供具體的行為範例，函式記憶則捕捉各函式的呼叫慣例與常見陷阱。工作流程記憶與子任務記憶在每個任務開始時被主動注入，而函式記憶則在工具呼叫發生錯誤時被反應式檢索。我們以 GPT-5-mini 作為教師模型，在三個工具使用基準上評估 AMD，使用四個學生模型（4B-8B 參數），在 AppWorld、BFCL V3 和 ToolSandbox 上分別達到平均 27.2 個百分點、11.2 個百分點和 3.4 個百分點的準確率提升，並持續優於現有的記憶式基線方法。進一步的分析顯示，子任務記憶貢獻了最大的提升幅度；教師的有效性取決於教師能力與學生相容性兩者；而 4B 規模的學生模型從 AMD 中獲益最多。

English

Memory systems have shown promise for improving agent performance, but their potential remains largely unexplored for small language models, which struggle to generate sufficient successful trajectories on their own. We propose Agent Memory Distillation (AMD), a training-free framework that transfers structured knowledge from a large teacher agent to a small student agent through hierarchical memory. AMD constructs three complementary memory types from successful teacher trajectories: Workflow memory encodes task-level strategies, Subtask memory provides concrete behavioral examples at an intermediate granularity, and Function memory captures per-function calling conventions and common pitfalls. Workflow and Subtask memories are injected proactively at the start of each task, while Function memory is retrieved reactively upon tool-calling errors. We evaluate AMD on three tool-use benchmarks using four student models (4B-8B parameters) with GPT-5-mini as the teacher, achieving average accuracy gains of 27.2%p, 11.2%p, and 3.4%p on AppWorld, BFCL V3, and ToolSandbox, while consistently outperforming existing memory-based baselines. Further analysis shows that Subtask memory contributes the largest gains, teacher effectiveness depends on both teacher capability and student compatibility, and 4B-sized students benefit most from AMD.