CLAG: Adaptive Memory Organization via Agent-Driven Clustering for Small Language Model Agents

Researchers introduce CLAG, a clustering-based memory framework that helps small language model agents organize and retrieve information more effectively. The system addresses memory dilution issues by creating semantic clusters with automated profiles, showing improved performance across multiple QA datasets.

• →CLAG framework enables small language models to actively organize memory through semantic clustering rather than single global pools.
• →The system uses an SLM-driven router to assign memories to coherent clusters with auto-generated topic summaries and tags.
• →Two-stage retrieval process first filters relevant clusters to reduce noise and improve search efficiency.
• →Experiments across multiple QA datasets demonstrate consistent improvements in answer quality and robustness.
• →The framework remains lightweight and efficient while addressing cross-topic interference in small language models.