Cost and accuracy of long-term memory in Distributed Multi-Agent Systems based on Large Language Models

Distributed multi-agent systems (DMAS) based on large language models (LLMs) enable collaborative intelligence while preserving data privacy. However, systematic evaluations of long-term memory under network constraints are limited. This study introduces a flexible testbed to compare mem0, a vector-based memory framework, and Graphiti, a graph-based knowledge graph, using the LoCoMo long-context benchmark. Experiments were conducted under unconstrained and constrained network conditions, measuring computational, financial, and accuracy metrics. Results indicate mem0 significantly outperforms Graphiti in efficiency, featuring faster loading times, lower resource consumption, and minimal network overhead. Crucially, accuracy differences were not statistically significant. Applying a statistical Pareto efficiency framework, mem0 is identified as the optimal choice, balancing cost and accuracy in DMAS.