受人类记忆启发，构建可长期维持的智能体记忆系统。
Human-Inspired Memory Architecture for LLM Agents
- 模拟睡眠巩固、遗忘、记忆再巩固等六种生物机制管理记忆。
- 在13K问题数据集上实现97.2%保留精度，存储量减少58%。
- 适用于需要长期记忆的对话与任务型AI系统。
当前大模型智能体缺乏在长时间交互中有效管理持久记忆的合理机制。本文提出一种基于生物学原理的记忆架构，包含六种认知机制：（1）睡眠期巩固，（2）干扰式遗忘，（3）痕迹成熟，（4）检索时的再巩固，（5）实体知识图谱，（6）混合多线索检索。每种机制针对单一记忆累积失效模式。我们引入一种合成校准方法，无需基准数据即可确定所有管道阈值，避免评估泄露。在两个基准上评估：第一，在包含13,000个问题和120,000条事件的VSCode问题跟踪数据集上，基于去重的巩固策略实现97.2%保留精度，存储量减少58%，优于基线21.8个百分点；第二，在LongMemEval个人对话基准上首次进行流式多层级评估（475会话，约54万独特交互）。在20万词符上下文预算下，本方案检索准确率70.1%（与原始检索71.2%无显著差异，95%置信区间重叠），同时提供可调的准确率/存储量权衡曲线。在S级规模（50会话）下，去重式巩固使偏好召回提升13.3个百分点。
原文摘要 · Abstract (English)
Current LLM agents lack principled mechanisms for managing persistent memory across long interaction horizons. We present a biologically-grounded memory architecture comprising six cognitive mechanisms: (1) sleep-phase consolidation, (2) interference-based forgetting, (3) engram maturation, (4) reconsolidation upon retrieval, (5) entity knowledge graphs, and (6) hybrid multi-cue retrieval. Each mechanism addresses a specific failure mode of naive memory accumulation. We introduce a synthetic calibration methodology that derives all pipeline thresholds without benchmark data exposure, eliminating a common source of evaluation leakage. We evaluate on two benchmarks. First, a VSCode issue-tracking dataset (13K issues, 120K events) where deduplication-based consolidation achieves 97.2% retention precision with 58% store reduction (+21.8 pp over baseline). Second, the LongMemEval personal-chat benchmark where we conduct the first streaming M-tier evaluation (475 sessions, ~540K unique turns). At a 200K-token context budget, our pipeline matches raw retrieval accuracy (70.1% vs. 71.2%, overlapping 95% CI) while exposing a tunable accuracy/store-size operating curve. At S-tier scale (50 sessions), dedup-based consolidation yields a +13.3 pp improvement in preference recall.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。