---
title: "AutoMem：将记忆作为可学习认知技能的自动化框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.01224v1"
permalink: "https://feed.astralor.com/items/itm_c36e5a21b5b7f24e405a75c8"
published: "2026-07-01T17:57:03.000Z"
score: 59
channel: "ai"
tags: ["AutoMem","LLM","记忆管理","长时程任务"]
---

# AutoMem：将记忆作为可学习认知技能的自动化框架

## 速览

AutoMem提出将记忆管理作为大型语言模型（LLM）的可训练技能，通过两个自动化循环优化记忆结构和记忆能力。第一个循环利用强LLM审查完整轨迹并迭代更新记忆结构，第二个循环从大量片段中提取好的记忆决策作为训练信号。在Crafter、MiniHack和NetHack三个长时程游戏中，仅优化记忆（不修改任务行为）就将基线性能提升约2-4倍，使32B开源模型媲美Claude Opus 4.5和Gemini 3.1 Pro Thinking。研究表明记忆管理是独立可学习的技能，对长时程任务具有高杠杆作用。

## 关键点

- 将文件系统操作提升为与任务动作同等的一等记忆动作，让模型自主决定记忆管理
- 双循环自动化优化：结构循环（强LLM审查轨迹改进记忆结构）和能力循环（从成功决策中训练记忆技能）
- 在三个程序化生成的长时程游戏中，仅优化记忆使性能提升2-4倍，32B模型达到前沿系统水平

## 全文

Memory expertise is a learned skill: knowing what to encode, when to retrieve, and how to organize knowledge--a capacity known in cognitive science as metamemory. We bring this perspective to LLMs by treating memory management as a trainable skill. We promote file-system operations to first-class memory actions alongside task actions, letting the model itself decide how to manage its memory. This memory skill improves along two axes: the structure that supports it (prompts, file schemas, action vocabulary), and the proficiency of the model exercising it. Both axes resist manual optimization: episodes in long-horizon tasks run for thousands of steps, and a single memory mistake can hide long before it surfaces, making human review of full trajectories impractical. We introduce AutoMem, a framework that automates both axes. In the first loop, a strong LLM reviews complete agent trajectories and iteratively revises the memory structure that shapes how the agent interacts with its memory files. In the second loop, the agent's own good memory decisions are identified from many episodes and used as training signal to sharpen the model's memory proficiency directly. Across three procedurally generated long-horizon games (Crafter, MiniHack, and NetHack), optimizing memory alone--without modifying the model's task-action behavior--improved the base agent's performance ~2x-4x, bringing a 32B open-weight model competitive with frontier systems such as Claude Opus 4.5 and Gemini 3.1 Pro Thinking. Our results show that memory management is an independently learnable skill, and a high-leverage objective yielding large gains on long-horizon tasks.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_c36e5a21b5b7f24e405a75c8