Trace Only What You Need: Structure-Aware On-Demand Hypergraph Memory for Long-Document Question Answering
Researchers have introduced DocTrace, a novel multi-agent retrieval-augmented generation (RAG) framework designed to enhance question answering over long documents. This system addresses limitations in existing RAG methods by organizing knowledge on-demand, leveraging document structure, and reusing past reasoning experiences. Experiments show DocTrace outperforms strong baselines on several datasets while significantly reducing computational costs. AI
IMPACT Enhances LLM reasoning over lengthy texts, potentially improving information retrieval and analysis in complex document sets.