Researchers have introduced $\Psi$-RAG, a novel framework designed to improve retrieval-augmented generation (RAG) for cross-document question answering. This new approach addresses limitations in existing tree-based RAG methods, such as structural isolation and coarse abstraction, by employing an iterative merging and collapse process to build a hierarchical abstract tree index. $\Psi$-RAG also features a multi-granular retrieval agent that enhances query processing and utilizes a hybrid retriever. The framework demonstrates significant performance gains, outperforming RAPTOR by 25.9% and HippoRAG 2 by 7.4% on cross-document multi-hop QA benchmarks. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Enhances cross-document retrieval for LLMs, potentially improving complex question-answering systems.
RANK_REASON This is a research paper introducing a new framework for retrieval-augmented generation.