发表机构
Birla Institute of Technology and Science, Pilani（比拉理工学院皮拉尼分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究借鉴灌丛鸦情景记忆原理，提出ScrubJay-MEM智能体记忆系统，通过类型条件时间衰减优化记忆管理，在TGT和MemoryAgentBench任务上取得优于现有方法的效果。
AI 中文摘要
跨会话持续存在的大语言模型（LLM）智能体积累的存储记忆，其有效性因内容类型差异极大，但现有记忆架构将所有记忆视为同等持久，导致检索上下文被过时事实系统性污染。我们表明，西灌丛鸦（western scrub jay）情景记忆的特性——按记忆类型分类的时间衰减，可被作为自动分类系数π_i在外部LLM智能体记忆存储中实现，由此产生ScrubJay-MEM：每条记忆被编码为联合绑定的“何事-何地-何时”（What--Where--When）元组，带有估计的易逝性π_i和效用窗口τ_i，通过查询自适应评分进行检索，每次更新时以O(1)次LLM调用进行回溯修正。我们引入时间泛化测试（Temporal Generalization Test, TGT），这是一个保留了未见过的保留间隔和泛化差距（Generalization Gap, GenGap）指标的基准。在TGT上，ScrubJay-MEM是唯一具有显著正GenGap（+0.108）的基于检索的系统；在MemoryAgentBench的EventQA-64k任务中，在某个LLM主干模型下，它的F1值比Mem0提升+2.66，比Qwen3-Embedding-4B提升+3.09。衰减消融实验使GenGap下降5.7倍，确立了类型条件衰减是该结果的必要条件。在更强的主干模型下，增益会缩小，而在事实整合任务上则会反转，将贡献范围限定于易逝事实的时间推理。
英文摘要
LLM agents that persist across sessions accumulate stored memories whose validity varies enormously by content type, yet existing memory architectures treat all memories as equally persistent and systematically contaminate retrieved context with outdated facts. We show that per-memory, type-conditioned temporal decay, a property of western scrub jay episodic memory, can be operationalized as an auto-classified coefficient $π_i$ in an external LLM-agent memory store, yielding ScrubJay-MEM: each memory is encoded as a jointly-bound What--Where--When tuple with an estimated perishability $π_i$ and utility horizon $τ_i$, retrieved by query-adaptive scoring, and revised retroactively at $O(1)$ LLM calls per update. We introduce the Temporal Generalization Test (TGT), a benchmark with held-out retention intervals and a Generalization Gap (GenGap) metric. On TGT, ScrubJay-MEM is the only retrieval-based system with substantially positive GenGap ($+0.108$); on MemoryAgentBench EventQA-64k it improves F1 by $+2.66$ over Mem0 and $+3.09$ over Qwen3-Embedding-4B under a llm backbone. A decay ablation collapses GenGap by $5.7\times$, establishing type-conditioned decay as necessary for the result. Gains narrow under stronger backbones and reverse on fact-consolidation tasks, scoping the contribution to temporal reasoning over perishable facts.
