relational-memory

MemPrism proposes a task-conditioned relational memory framework that separates persistent experience storage from decision-time working memory, enabling long-horizon agents to dynamically construct relational views for improved performance and reduced token usage.

SubtleMemory是一个用于评估AI代理在长期交互中细粒度关系记忆辨别能力的基准，包含10个长历史中的1,522个实例。它揭示了当前记忆系统在保存和利用细微记忆关系方面的局限性。

HAGE引入了一种加权多关系记忆框架，能够在统一关系记忆图上实现查询条件化的遍历，通过自适应记忆检索和基于强化学习的优化来提高长程推理准确性。