RAG

Hierarchical RAG framework that compresses and partitions documents at multiple scales to address imprecise retrieval and fragmented context in long-context LLMs