dorsal/arxiv
View SchemaHiMeS: Hippocampus-inspired Memory System for Personalized AI Assistants
| Authors | Hailong Li, Feifei Li, Wenhui Que, Xingyu Fan |
|---|---|
| Categories | |
| ArXiv ID | 2601.06152vv1 |
| URL | https://arxiv.org/abs/2601.06152 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Large language models (LLMs) power many interactive systems such as chatbots, customer-service agents, and personal assistants. In knowledge-intensive scenarios requiring user-specific personalization, conventional retrieval-augmented generation (RAG) pipelines exhibit limited memory capacity and insufficient coordination between retrieval mechanisms and user-specific conversational history, leading to redundant clarification, irrelevant documents, and degraded user experience. Inspired by the hippocampus-neocortex memory mechanism, we propose HiMeS, an AI-assistant architecture that fuses short-term and long-term memory. Our contributions are fourfold: (1) A short-term memory extractor is trained end-to-end with reinforcement learning to compress recent dialogue and proactively pre-retrieve documents from the knowledge base, emulating the cooperative interaction between the hippocampus and prefrontal cortex. (2) A partitioned long-term memory network stores user-specific information and re-ranks retrieved documents, simulating distributed cortical storage and memory reactivation. (3) On a real-world industrial dataset, HiMeS significantly outperforms a cascaded RAG baseline on question-answering quality. (4) Ablation studies confirm the necessity of both memory modules and suggest a practical path toward more reliable, context-aware, user-customized LLM-based assistants.
{
"annotation_id": "a2a2d384-f66a-48db-a1f7-79558702e2c4",
"date_created": "2026-02-17T05:53:05.007000Z",
"date_modified": "2026-02-17T05:53:05.007000Z",
"file_hash": "289e5ee1bdc739f967a170f749112015147bb3360af7e43c2c00ee7c14c0f960",
"private": false,
"record": {
"abstract": "Large language models (LLMs) power many interactive systems such as chatbots, customer-service agents, and personal assistants. In knowledge-intensive scenarios requiring user-specific personalization, conventional retrieval-augmented generation (RAG) pipelines exhibit limited memory capacity and insufficient coordination between retrieval mechanisms and user-specific conversational history, leading to redundant clarification, irrelevant documents, and degraded user experience. Inspired by the hippocampus-neocortex memory mechanism, we propose HiMeS, an AI-assistant architecture that fuses short-term and long-term memory. Our contributions are fourfold: (1) A short-term memory extractor is trained end-to-end with reinforcement learning to compress recent dialogue and proactively pre-retrieve documents from the knowledge base, emulating the cooperative interaction between the hippocampus and prefrontal cortex. (2) A partitioned long-term memory network stores user-specific information and re-ranks retrieved documents, simulating distributed cortical storage and memory reactivation. (3) On a real-world industrial dataset, HiMeS significantly outperforms a cascaded RAG baseline on question-answering quality. (4) Ablation studies confirm the necessity of both memory modules and suggest a practical path toward more reliable, context-aware, user-customized LLM-based assistants.",
"arxiv_id": "2601.06152",
"authors": [
"Hailong Li",
"Feifei Li",
"Wenhui Que",
"Xingyu Fan"
],
"categories": [
"cs.AI"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "HiMeS: Hippocampus-inspired Memory System for Personalized AI Assistants",
"url": "https://arxiv.org/abs/2601.06152",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "399850e6-e0d0-4ae6-b92e-9e0a92da208c",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}