Tag
This paper proposes a Dual-Layer Agentic Memory framework for LLM agents that manages knowledge lifecycle through fast write routing and slow consolidation to prune redundant external memory while maintaining performance.