---
id: 20261104-T0-04
title: "MEMENTO：让长上下文智能体学会管理自己的记忆"
title_en: "MEMENTO: Teaching Long-Context Agents to Manage Their Own Memory"
url: https://ai.daily.yangsir.net/daily/20261104-T0-04
issue_date: 2026-11-04
publish_date: 2026-11-03T05:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2611.00244
---

# MEMENTO：让长上下文智能体学会管理自己的记忆

arXiv 新论文 MEMENTO 研究长输入场景下 LLM 智能体的上下文管理问题：在固定大小的上下文窗口里，哪些内容该保留。常见做法是压缩，即模型用自己生成的笔记替换历史记录，但现成的智能体方案在这上面表现不佳。MEMENTO 提出学习式的上下文管理方法。对需要处理长文档或多轮任务的智能体来说，记忆取舍策略直接决定任务成败。

## English Version

**MEMENTO: Teaching Long-Context Agents to Manage Their Own Memory**

New arXiv paper MEMENTO addresses context management for LLM agents on long inputs: deciding what to retain within a fixed-size context window. A common approach is compaction, where the model replaces its history with a self-generated note, but off-the-shelf agentic methods perform poorly at this. MEMENTO proposes a learned approach to context management. For agents handling long documents or multi-turn tasks, memory retention strategy directly determines task success.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2611.00244)

**详情页**：https://ai.daily.yangsir.net/daily/20261104-T0-04

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*