核心信息
RoMem 是一个可即插即用的重排序器,用来修复智能体对“过时事实”的静默误记:它不会删除旧信息,而是在查询时让过时事实旋转出对齐范围,使智能体优先使用最新知识。
要点
- 即插即用:以重排序器形式接入,大约三行代码即可集成。
- 旋转而非删除:过时事实不会被清除,只会在查询时被“遮到后面”。
- 关系时效感知:预训练门控会估算不同关系的变化速度(如 born_in 几乎不变、works_at 变化很快),并可零样本推广到未见关系。
- 把时间当作旋转:用连续旋转建模时间,而不是简单打一个时间标签。
RoMem 是一个可即插即用的重排序器,用来修复智能体对“过时事实”的静默误记:它不会删除旧信息,而是在查询时让过时事实旋转出对齐范围,使智能体优先使用最新知识。
Fun fact: your agent is quietly wrong about the past. Outdated facts aren't deleted, just shadowed to the back. RoMem fixes that. Drop-in reranker, three lines to plug in. The key ideas: > Spin, don't delete Outdated facts rotate out of alignment at query time. Nothing erased. > Some relations expire faster A pretrained gate scores how fast each relation changes. born_in → near zero. works_at → fast. Zero-shot to unseen relations. > Time as rotation, not a label One continuous function. Any timestamp. Including dates never seen in training. 2–3x better at temporal reasoning, with zero drop in static memory. Joint research from Edinburgh, LIGHTSPEED and St Andrews. EMNLP 2026 Main. Paper: https://t.co/0tt4nFDhM6 Code: https://t.co/VFPUKlBRuO