Scaling Teams or Scaling Time? Memory Enabled Lifelong Learning in LLM Multi-Agent Systems

Researchers introduce LLMA-Mem, a memory framework for LLM multi-agent systems that balances team size with lifelong learning capabilities. The study reveals that larger agent teams don't always perform better long-term, and smaller teams with better memory design can outperform larger ones while reducing costs.

• →LLMA-Mem framework enables LLM multi-agent systems to learn and improve through accumulated experience over time.
• →Larger agent teams do not always produce better long-term performance compared to smaller, well-designed teams.
• →Memory design is identified as a practical path for scaling multi-agent systems more effectively and efficiently.
• →The framework consistently improves long-horizon performance while reducing operational costs.
• →Non-monotonic scaling reveals that optimal team size depends on memory architecture and experience reuse capabilities.