dorsal/arxiv
View SchemaSwiftMem: Fast Agentic Memory via Query-aware Indexing
| Authors | Anxin Tian, Yiming Li, Xing Li, Hui-Ling Zhen, Lei Chen, Xianzhi Yu, Zhenhua Dong, Mingxuan Yuan |
|---|---|
| Categories | |
| ArXiv ID | 2601.08160vv1 |
| URL | https://arxiv.org/abs/2601.08160 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Agentic memory systems have become critical for enabling LLM agents to maintain long-term context and retrieve relevant information efficiently. However, existing memory frameworks suffer from a fundamental limitation: they perform exhaustive retrieval across the entire storage layer regardless of query characteristics. This brute-force approach creates severe latency bottlenecks as memory grows, hindering real-time agent interactions. We propose SwiftMem, a query-aware agentic memory system that achieves sub-linear retrieval through specialized indexing over temporal and semantic dimensions. Our temporal index enables logarithmic-time range queries for time-sensitive retrieval, while the semantic DAG-Tag index maps queries to relevant topics through hierarchical tag structures. To address memory fragmentation during growth, we introduce an embedding-tag co-consolidation mechanism that reorganizes storage based on semantic clusters to improve cache locality. Experiments on LoCoMo and LongMemEval benchmarks demonstrate that SwiftMem achieves 47$\times$ faster search compared to state-of-the-art baselines while maintaining competitive accuracy, enabling practical deployment of memory-augmented LLM agents.
{
"annotation_id": "d851604c-b072-4036-9a40-b7c4f95c95c2",
"date_created": "2026-02-17T05:53:15.098000Z",
"date_modified": "2026-02-17T05:53:15.098000Z",
"file_hash": "eeeca881a88f467417cfbf590960e33340ad3dcbdf07cf49e6120f37e8b68cde",
"private": false,
"record": {
"abstract": "Agentic memory systems have become critical for enabling LLM agents to maintain long-term context and retrieve relevant information efficiently. However, existing memory frameworks suffer from a fundamental limitation: they perform exhaustive retrieval across the entire storage layer regardless of query characteristics. This brute-force approach creates severe latency bottlenecks as memory grows, hindering real-time agent interactions. We propose SwiftMem, a query-aware agentic memory system that achieves sub-linear retrieval through specialized indexing over temporal and semantic dimensions. Our temporal index enables logarithmic-time range queries for time-sensitive retrieval, while the semantic DAG-Tag index maps queries to relevant topics through hierarchical tag structures. To address memory fragmentation during growth, we introduce an embedding-tag co-consolidation mechanism that reorganizes storage based on semantic clusters to improve cache locality. Experiments on LoCoMo and LongMemEval benchmarks demonstrate that SwiftMem achieves 47$\\times$ faster search compared to state-of-the-art baselines while maintaining competitive accuracy, enabling practical deployment of memory-augmented LLM agents.",
"arxiv_id": "2601.08160",
"authors": [
"Anxin Tian",
"Yiming Li",
"Xing Li",
"Hui-Ling Zhen",
"Lei Chen",
"Xianzhi Yu",
"Zhenhua Dong",
"Mingxuan Yuan"
],
"categories": [
"cs.CL",
"cs.AI"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "SwiftMem: Fast Agentic Memory via Query-aware Indexing",
"url": "https://arxiv.org/abs/2601.08160",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "0aec75cd-0141-4741-9717-3d1e306451cd",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}