时间非标签：面向时序知识图谱与智能体记忆的连续相位旋转机制

Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory

April 13, 2026

作者: Weixian Waylon Li, Jiaxin Zhang, Xianan Jim Yang, Tiejun Ma, Yiwen Guo

cs.AI

摘要

诸如知识图谱之类的结构化记忆表征是自主智能体及其他长效系统的核心。然而现有方法大多将时间建模为离散元数据——或按时效性排序（导致陈旧但永久的知识被埋没），或直接覆盖过时事实，或在每次信息录入时都需调用成本高昂的大语言模型，因而无法区分持久性事实与演变性事实。为此，我们提出RoMem：一种即插即用的时序知识图谱模块，适用于智能体记忆等结构化记忆系统。该模块通过预训练的语义速度门控，将关系文本嵌入映射为波动分数，从数据中学习到演变关系（如"某国总统"）应快速轮转，而持久关系（如"出生地"）应保持稳定。结合连续相位旋转机制，实现了几何遮蔽：过时事实在复数向量空间中被相位旋转，使得时序正确的事实能自然压制矛盾陈述而无需删除操作。在时序知识图谱补全任务中，RoMem在ICEWS05-15数据集上达到72.6的MRR值，刷新业界最优纪录。应用于智能体记忆时，其在时序推理任务（MultiTQ）上实现MRR与答案准确率2-3倍提升，主导混合基准测试（LoCoMo），静态记忆保持零衰减（DMR-MSC），并能零样本泛化至未知金融领域（FinTMMBench）。

English

Structured memory representations such as knowledge graphs are central to autonomous agents and other long-lived systems. However, most existing approaches model time as discrete metadata, either sorting by recency (burying old-yet-permanent knowledge), simply overwriting outdated facts, or requiring an expensive LLM call at every ingestion step, leaving them unable to distinguish persistent facts from evolving ones. To address this, we introduce RoMem, a drop-in temporal knowledge graph module for structured memory systems, applicable to agentic memory and beyond. A pretrained Semantic Speed Gate maps each relation's text embedding to a volatility score, learning from data that evolving relations (e.g., "president of") should rotate fast while persistent ones (e.g., "born in") should remain stable. Combined with continuous phase rotation, this enables geometric shadowing: obsolete facts are rotated out of phase in complex vector space, so temporally correct facts naturally outrank contradictions without deletion. On temporal knowledge graph completion, RoMem achieves state-of-the-art results on ICEWS05-15 (72.6 MRR). Applied to agentic memory, it delivers 2-3x MRR and answer accuracy on temporal reasoning (MultiTQ), dominates hybrid benchmark (LoCoMo), preserves static memory with zero degradation (DMR-MSC), and generalises zero-shot to unseen financial domains (FinTMMBench).