HiMem: Hierarchical Long-Term Memory for LLM Long-Horizon Agents

πŸ“… 2026-01-10
πŸ›οΈ arXiv.org
πŸ“ˆ Citations: 1
✨ Influential: 0
πŸ“„ PDF
πŸ€– AI Summary
Existing long-term memory systems struggle to simultaneously achieve adaptability, scalability, and self-evolution in continuous interaction scenarios. To address this challenge, this work proposes HiMemβ€”a cognitively inspired hierarchical long-term memory framework that dynamically constructs, retrieves, and updates memories by establishing semantic associations between event memory and note memory. The core innovations include a topic-aware dual-channel segmentation strategy that distinguishes routine events from unexpected ones, and a conflict-aware memory reconsolidation mechanism, both integrated with multi-stage information extraction and a hybrid retrieval strategy. Experimental results demonstrate that HiMem significantly outperforms current state-of-the-art methods on long-term dialogue benchmarks, achieving leading performance in accuracy, consistency, long-range reasoning capability, and computational efficiency.

Technology Category

Application Category

πŸ“ Abstract
Although long-term memory systems have made substantial progress in recent years, they still exhibit clear limitations in adaptability, scalability, and self-evolution under continuous interaction settings. Inspired by cognitive theories, we propose HiMem, a hierarchical long-term memory framework for long-horizon dialogues, designed to support memory construction, retrieval, and dynamic updating during sustained interactions. HiMem constructs cognitively consistent Episode Memory via a Topic-Aware Event--Surprise Dual-Channel Segmentation strategy, and builds Note Memory that captures stable knowledge through a multi-stage information extraction pipeline. These two memory types are semantically linked to form a hierarchical structure that bridges concrete interaction events and abstract knowledge, enabling efficient retrieval without sacrificing information fidelity. HiMem supports both hybrid and best-effort retrieval strategies to balance accuracy and efficiency, and incorporates conflict-aware Memory Reconsolidation to revise and supplement stored knowledge based on retrieval feedback. This design enables continual memory self-evolution over long-term use. Experimental results on long-horizon dialogue benchmarks demonstrate that HiMem consistently outperforms representative baselines in accuracy, consistency, and long-term reasoning, while maintaining favorable efficiency. Overall, HiMem provides a principled and scalable design paradigm for building adaptive and self-evolving LLM-based conversational agents. The code is available at https://github.com/jojopdq/HiMem.
Problem

Research questions and friction points this paper is trying to address.

long-term memory
LLM agents
adaptability
scalability
self-evolution
Innovation

Methods, ideas, or system contributions that make the work stand out.

Hierarchical Memory
Long-Horizon Dialogue
Memory Reconsolidation
Event-Surprise Segmentation
Self-Evolving Agents
N
Ningning Zhang
Macau University of Science and Technology
X
Xingxing Yang
Macau University of Science and Technology
Z
Zhizhong Tan
Macau University of Science and Technology
W
Weiping Deng
Macau University of Science and Technology
W
Wenyong Wang
Macau University of Science and Technology