发表机构
Technical University of Munich; inovex GmbH; Cerebras Systems Inc.（慕尼黑工业大学; inovex有限公司; Cerebras系统公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
MemoryLACE是一种轻量级记忆框架，通过显式建模文本证据的生命周期，在两个数据集上实现同骨干模型下的最高性能，且运行时间大幅减少，提升了LLM智能体的长期记忆推理能力。
AI 中文摘要
长期大语言模型（LLM）智能体必须在多轮交互中保留信息，同时区分重复证据、历史状态、更新内容及未解决的矛盾。现有文本记忆系统能高效检索语义相关记忆，但常将这些关系隐含处理；而更丰富的结构化方法则通过全局图、层次抽象或反思建模这些关系，复杂度更高。我们提出MemoryLACE（简称MemLACE），这是一种轻量级记忆框架，通过稀疏合并、替代和矛盾关系显式建模文本证据的生命周期，同时保留原子自然语言记忆及其来源。MemLACE并非独立检索记忆，而是重构感知关系的证据单元，为下游推理暴露当前、历史、支持性及冲突性证据。在BEAM和StructMemEval数据集上，使用开放权重及专有LLM骨干模型，MemLACE在同骨干模型比较中实现最高整体性能，且在BEAM上的端到端运行时间比最强的反思记忆基线Hindsight减少66.6%。 ablation研究表明，生命周期扩展和时间感知是这些提升的主要贡献因素。综合结果显示，显式建模文本证据的局部生命周期足以大幅改善长期记忆推理，无需依赖全面知识图谱或全局反思。
英文摘要
Long-term LLM agents must preserve information across interactions while distinguishing repeated evidence, historical states, updates, and unresolved contradictions. Existing textual memory systems retrieve semantically relevant memories efficiently but often leave these relationships implicit, whereas richer structured approaches model them through global graphs, hierarchical abstractions, or reflection at greater complexity. We introduce MemoryLACE (MemLACE), a lightweight memory framework that explicitly models the lifecycle of textual evidence through sparse merge, supersession, and contradiction relations while preserving atomic natural-language memories and their provenance. Rather than retrieving memories independently, MemLACE reconstructs relation-aware evidence units that expose current, historical, supporting, and conflicting evidence for downstream reasoning. Across BEAM and StructMemEval, using open-weight and proprietary LLM backbones, MemLACE achieves the highest overall performance in same-backbone comparisons while reducing end-to-end runtime on BEAM by 66.6% relative to Hindsight, the strongest reported reflective-memory baseline. Ablation studies identify lifecycle expansion and temporal awareness as the principal contributors to these gains. Together, the results demonstrate that explicitly modeling the local lifecycle of textual evidence is sufficient to substantially improve long-term memory reasoning without requiring comprehensive knowledge graphs or global reflection.
Comments8 pages, 2 figures, 4 tables
