← Home首页
Open Research

Research

研究

Open research outputs from building long-running agent memory systems. Independent work, published as it matures. Part of OpenSoul.

来自长期运行的智能体记忆系统建设的公开研究成果。独立研究,成熟即公开。属于 OpenSoul。

Outputs成果

1
Preprint 2026-10

Salience, Ranking, and Metabolism: Three Conflated Signals in Long-Running Agent Memory Systems

Baofeng Zhao · Independent Researcher

Agent memory treats “importance” as one signal. We show it is three conflated ones — salience, ranking, and metabolism — and document the separation with production failure evidence from about 7 months of a live system.

智能体记忆把“重要性”当成一个信号。我们论证它其实是三个被混用的信号——浓度、排序、代谢——并用一个真实系统自 2026-03 起约 7 个月的生产故障证据记录它们的分离。

More outputs in preparation: further notes on agent memory and the audit tools described in the paper. 更多成果筹备中:关于智能体记忆的后续笔记,以及论文所述的自查工具。