研究库 论文知识库

内容库 / 论文

arXiv:2610.10533 · RAG 检索增强

EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory

EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory

类型：arxiv
标识：2610.10533
链接：https://arxiv.org/abs/2610.10533
主分类：rag
形态：method
TLDR：Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings c
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-10-08-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记 Cool Papers · cs.CL (papers.cool) · RSS 摘要

仓库 oigrex666/mcp-context-chronicler

笔记 Tom 文献雷达 · Agent · RAG · Long Context · 2026-09-13 20:40

仓库 arunksn/Adaptive-Conversational-Memory

笔记 Cool Papers · cs.IR (papers.cool) · RSS 摘要

笔记 flyP 短审稿 · Think Before You Link（Rarity, Reasoning, and Retrieval in Multilingual Entity Linking）

arXiv

2610.10533

TLDR

Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings c

arXiv ↗ 中文全文