HIRO: Hierarchical Information Retrieval Optimization

TL;DR

HIRO is a novel hierarchical querying method for Retrieval-Augmented Generation that reduces information overload in LLMs, improving performance by 10.85% on NarrativeQA through DFS-based recursive scoring and branch pruning.

Contribution

HIRO introduces a depth-first search-based recursive similarity scoring and branch pruning technique to optimize hierarchical data retrieval for LLMs.

Findings

Achieved 10.85% performance improvement on NarrativeQA.

Effectively manages information overload in hierarchical RAG systems.

Minimizes context size without losing essential information.

Abstract

Retrieval-Augmented Generation (RAG) has revolutionized natural language processing by dynamically integrating external knowledge into Large Language Models (LLMs), addressing their limitation of static training datasets. Recent implementations of RAG leverage hierarchical data structures, which organize documents at various levels of summarization and information density. This complexity, however, can cause LLMs to "choke" on information overload, necessitating more sophisticated querying mechanisms. In this context, we introduce Hierarchical Information Retrieval Optimization (HIRO), a novel querying approach that employs a Depth-First Search (DFS)-based recursive similarity score calculation and branch pruning. This method uniquely minimizes the context delivered to the LLM without informational loss, effectively managing the challenge of excessive data. HIRO's refined approach is validated by a 10.85% improvement in performance on the NarrativeQA dataset.