# Awesome-Memory-for-Agents — Paper List + Taxonomy

> Awesome-Memory-for-Agents is a paper list and taxonomy for agent memory, splitting short vs long-term memory and mapping to 3 application scenarios.

## Install

Copy the content below into your project:

## Quick Use

```bash
git clone https://github.com/tsinghuac3i/awesome-memory-for-agents.git
cd awesome-memory-for-agents
rg -n "## Overview|## Benchmark" README.md
# open README.md and follow paper links
```

## Intro

Awesome-Memory-for-Agents is a paper list and taxonomy for agent memory, splitting short vs long-term memory and mapping to 3 application scenarios.

**Best for:** agent builders who want a structured reading list for memory systems

**Works with:** Git + Markdown viewer; link out to arXiv papers

**Setup time:** 2-5 minutes

### Key facts (verified)

- GitHub: 487 stars · 34 forks · pushed 2026-05-13.
- License: MIT · owner avatar + repo URL verified via GitHub API.
- README-verified entrypoint: `git clone https://github.com/tsinghuac3i/awesome-memory-for-agents.git`.

## Main

- Use the taxonomy first: it defines short-term vs long-term memory, then splits long-term into Experience vs Memory based on outcome validation.

- Pick one of the three application scenarios (personalization, learning from experience, long-horizon tasks) and skim the newest papers first.

- Use it as a design checklist: map your agent’s memory store (scratchpad, episodic log, external DB, skill store) to the repo’s terms to avoid ambiguity.

### Source-backed notes

- README’s Overview section defines Short-Term Memory vs Long-Term Memory, and further splits long-term memory into Experience vs Memory.
- README maps the taxonomy to three application scenarios: personalization, learning from experience, and long-horizon agentic tasks.
- The paper list is presented as dated tables with direct paper links (e.g., arXiv).

### FAQ

- **Is this a tool or a reading list?**: It’s a curated paper list; use it to guide design and evaluation of memory systems.
- **How should I read it efficiently?**: Start with one application scenario, then scan the newest dated entries first.
- **Does it include benchmarks?**: Yes — README includes a Benchmark section and organizes papers around application needs.

## Source & Thanks

> Source: https://github.com/TsinghuaC3I/Awesome-Memory-for-Agents
> License: MIT
> GitHub stars: 487 · forks: 34

---

<!-- ZH -->

## Quick Use

```bash
git clone https://github.com/tsinghuac3i/awesome-memory-for-agents.git
cd awesome-memory-for-agents
rg -n "## Overview|## Benchmark" README.md
# open README.md and follow paper links
```

## Intro

Awesome-Memory-for-Agents 汇总 Agent Memory 论文并给出分类体系：短期/长期记忆、Experience vs Memory，并按个性化、经验学习、长跨度任务三类场景组织阅读与 Benchmark。

**Best for:** 想系统性补齐 agent memory 研究脉络与术语的人

**Works with:** Git + Markdown 阅读器；每条条目链接到 arXiv/论文页

**Setup time:** 2-5 minutes

### Key facts (verified)

- GitHub：487 stars · 34 forks；最近更新 2026-05-13。
- 许可证：MIT；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中核对过的入口命令：`git clone https://github.com/tsinghuac3i/awesome-memory-for-agents.git`。

## Main

- 先看 taxonomy：它先区分短期/长期记忆，再把长期记忆按是否依赖任务结果验证分成 Experience 与 Memory。

- 按三类应用场景选路径（个性化、经验学习、长跨度任务），优先浏览最新条目再回溯经典工作。

- 把它当设计检查表：将你的 agent 记忆实现（scratchpad、日志、外部 DB、技能库）映射到术语体系，避免概念混用。

### Source-backed notes

- README 的 Overview 定义了短期记忆与长期记忆，并进一步区分 Experience 与 Memory。
- README 把 taxonomy 映射到三类应用场景：个性化、经验学习、长跨度 agentic 任务。
- 论文清单以带日期的表格呈现，并为每条提供论文链接（如 arXiv）。

### FAQ

- **这是工具还是论文清单？**：它是论文清单与分类体系，主要用于指导 memory 系统设计与评估。
- **怎么高效阅读？**：先选一个应用场景，再按日期从新到旧快速扫一遍，最后精读关键论文。
- **包含 Benchmark 吗？**：包含。README 有 Benchmark 章节，并围绕应用场景组织相关工作。

## Source & Thanks

> Source: https://github.com/TsinghuaC3I/Awesome-Memory-for-Agents
> License: MIT
> GitHub stars: 487 · forks: 34


---
Source: https://tokrepo.com/en/workflows/awesome-memory-for-agents-paper-list-taxonomy
Author: AI Open Source