MemForest: Efficient Agent Memory Management via EventTree Partitioning and Progressive Merging
Junxi Wang and collaborators across Shanghai Jiao Tong University, Fudan, Nanjing University, HIT and Sichuan University present MemForest, a memory compression layer that partitions history into event units, merges redundant nodes along a maximum spanning tree, and retrieves by propagating from anchor nodes.
Ask this paper
Event-centric partitioning: History is split into units using global semantic similarity together with local temporal continuity, so an event stays intact rather than being cut at an arbitrary token boundary.
EventTree merging: Each unit gets a maximum spanning tree, and redundant memory nodes are merged progressively by selecting high-weight edges, which is where the storage reduction comes from.
Anchor-guided propagation retrieval: Retrieval pulls relevant nodes from the temporal neighborhoods of key nodes rather than by similarity alone, which recovers accuracy lost to compression.
Numbers under Mem0: At 50% compression across LoCoMo, LongMemEval and PersonaMem, MemForest retains 97.1% of original performance with a 1.89x retrieval speedup.
Numbers under M3-Agent: On the multimodal M3-Agent framework it preserves 99.7% of performance at the same 50% compression with a 2.24x speedup, and it is framework-agnostic by design.
Abstract
Agent memory systems have demonstrated significant potential in long-term dialogue, personalized assistants, and video understanding. However, continuously accumulated memory introduces substantial storage and retrieval costs during inference. To address this issue, we propose \textbf{MemForest}, a general memory compression framework adaptable to various agent memory systems. Specifically, MemForest partitions historical memory into event-centric units by leveraging global semantic similarity and local temporal continuity. For each unit, it constructs a maximum spanning tree, termed an EventTree, and progressively merges redundant memory nodes by selecting high-weight edges, reducing storage overhead. Furthermore, we introduce an anchor-guided propagation retrieval mechanism that retrieves relevant memory nodes from the temporal neighborhoods of key nodes, improving retrieval accuracy. Extensive experiments demonstrate the effectiveness of MemForest. Under the unimodal Mem0 framework, MemForest retains \textbf{97.1%} of the original performance while compressing \textbf{50%} of historical memory across three benchmarks (LoCoMo, LongMemEval, and PersonaMem), achieving a \textbf{1.89x} retrieval speedup. Under the multimodal M3-Agent framework, it preserves \textbf{99.7%} of the original performance with a \textbf{50%} compression ratio across two benchmarks (M3-Bench-robot and M3-Bench-web), achieving a \textbf{2.24x} retrieval speedup. \textcolor{RoyalBlue}{\textit{Our code is available at [https://github.com/Celina-love-sweet/MemForest.}}](https://github.com/Celina-love-sweet/MemForest.}})