PAPER / ARXIV:2609.08273
Junxi Wang, Te Sun, Jiayi Zhu, Chen Zhang, Siyuan Li, Xuyang Liu, Zichen Wen, Xiaobing Tu, Jinkui Ren, Xiantao Zhang, Ziqi Yuan, Linfeng Zhang
RESUMO
Agent memory systems have demonstrated significant potential in long-term dialogue, personalized assistants, and video understanding, but continuously accumulated memory introduces substantial storage and retrieval costs. We propose MemForest, a general memory compression framework that partitions historical memory into event-centric units, constructs an EventTree per unit, and progressively merges redundant memory nodes. Under the Mem0 framework, MemForest retains 97.1% of original performance while compressing 50% of historical memory, achieving a 1.89x retrieval speedup.
NO MESMO MAPA