Zero-Mem：零Token的LLM Agent记忆管理方案
What if an AI agent could manage long-term memory without spending a single LLM…
What if an AI agent could manage long-term memory without spending a single LLM token on the memory itself?
如果AI代理能够在不花费任何LLM令牌的情况下管理长期记忆，那会怎样？
This paper shows an Agent memory may not need an LLM doing extra work every time it remembers something.
本文表明，代理的记忆可能不需要LLM在每次记忆时做额外的工作。
Zero-Mem’s memory operations used zero LLM tokens and cut latency by 57.6% versus the fastest baseline in the authors’ controlled comparison.
在作者的对照实验中，Zero-Mem的记忆操作使用了零个LLM令牌，并且与最快的基线相比，延迟降低了57.6%。
Zero-Mem keeps the original interaction history intact, then builds two non-generative views over it: an entity-context graph for relationships and a temporal hierarchy for session and local context.
Zero-Mem保持原始交互历史不变，然后在其上构建两个非生成视图：一个用于关系的实体上下文图和一个用于会话和局部上下文的时间层次结构。
At query time, deterministic routing decides how much to rely on each view, retrieves evidence, fills in relational or nearby context, and calibrates the result before the final answer model sees it.
在查询时，确定性路由决定每个视图的依赖程度，检索证据，填充关系或邻近上下文，并在最终答案模型看到结果之前进行校准。
That means every memory operation outside final QA uses zero LLM calls and zero LLM input or output tokens.
这意味着最终QA之外的所有记忆操作都使用零LLM调用和零LLM输入或输出令牌。
Structured agent memory can stay faithful to raw traces without turning memory management into a recurring generation problem.
结构化代理记忆可以忠实于原始痕迹，而无需将记忆管理变成反复出现的生成问题。
– arxiv. org/abs/2607.29377
– arxiv.org/abs/2607.29377
Title: "Zero-Mem: Zero-Token Memory Operations for LLM Agents"
标题：“Zero-Mem：LLM代理的零令牌记忆操作”
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力