---
id: 20260429-T0-03
title: "Memanto：基于信息论检索的 typed 语义内存系统"
title_en: "Memanto: Typed Semantic Memory with Information-Theoretic Retrieval"
url: https://ai.daily.yangsir.net/daily/20260429-T0-03
issue_date: 2026-04-29
publish_date: 2026-04-28T04:00:00.000Z
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2604.22085
---

# Memanto：基于信息论检索的 typed 语义内存系统

arXiv论文提出Memanto系统，解决了长周期智能体的内存架构瓶颈。该系统使用信息论检索方法管理typed语义内存，提高了持久多会话自主代理的效率。研究指出现有方法在状态模型推理到持久代理的过渡中存在内存处理不足的问题，Memanto通过信息论方法优化了检索过程。

## English Version

**Memanto: Typed Semantic Memory with Information-Theoretic Retrieval**

ArXiv paper introduces Memanto, addressing memory bottlenecks in long-horizon agents. The system uses information-theoretic retrieval for typed semantic memory, improving persistent multi-session autonomous agents. Research critiques existing methods' memory limitations during stateful to persistent agent transitions.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2604.22085)

**详情页**：https://ai.daily.yangsir.net/daily/20260429-T0-03

---

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