做 AI 记忆系统和长期上下文管理的开发者，这个算法更新直接解决了记忆时效性和检索效率的痛点，值得关注并尝试集成。
Mem0 发布了四月算法更新，引入了单次提取和分层检索机制，显著提升了记忆提取效率并降低了全上下文 token 成本。新算法包含时间推理功能，为每条记忆赋予时间戳，记录事件发生时间、是否持续或已完成、时间精度及记忆类型。同时增加了记忆衰减机制，基于时效性进行排序，确保旧记忆不会主导当前查询结果。这些改进使检索更具时间感知能力，能准确反映信息的变化和当前有效性。
What changed: The April algorithm shipped with single-pass extraction and hierarchical retrieval: ...
What changed: The April algorithm shipped with single-pass extraction and hierarchical retrieval: extract memories efficiently, retrieve across multiple signals, and stay far below full-context token costs. The updated algorithm has: 1. Temporal Reasoning- gives every memory a time signature: when the event happened, whether it is ongoing or completed, how precise the timing is, and what kind of memory it is. 2. Memory Decay- adds recency-aware ranking, so stale memories stay available without dominating current queries. Together, they make retrieval more time-aware: what changed, what is still true, and what should no longer dominate the result set. 💬 1 🔄 1 ❤️ 10 👀 863 📊 3 ⚡