---
id: 20260923-T0-19
title: "AdaMem：RAG软压缩中的自适应记忆token分配方案"
title_en: "AdaMem: Adaptive Memory Token Allocation for RAG Soft Compression"
url: https://ai.daily.yangsir.net/daily/20260923-T0-19
issue_date: 2026-09-23
publish_date: 2026-09-22T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2609.22100
---

# AdaMem：RAG软压缩中的自适应记忆token分配方案

arXiv新论文提出AdaMem，针对检索增强生成（RAG）中处理大量长段落成本高且易引入干扰信息的问题。该方法通过软压缩将段落编码为紧凑序列，并自适应分配记忆token，以在压缩率和信息保留之间取得更好平衡。AdaMem旨在降低RAG的推理开销，同时减少无关内容对生成质量的干扰，对需要处理长文档的RAG应用有参考价值。

## English Version

**AdaMem: Adaptive Memory Token Allocation for RAG Soft Compression**

A new arXiv paper introduces AdaMem for soft compression in retrieval-augmented generation (RAG). Processing many long passages is costly and can introduce distracting information. AdaMem encodes passages as compact sequences and adaptively allocates memory tokens to better balance compression rate and information retention. It aims to reduce RAG inference overhead while minimizing irrelevant content, offering value for RAG applications handling long documents.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2609.22100)

**详情页**：https://ai.daily.yangsir.net/daily/20260923-T0-19

---

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