---
format: "aidr-story-markdown/v1"
id: "5acdf59aa24dc1886413a47e7a8d43a71174c685118c6732e93d20e70ae6fbc3"
canonical_url: "https://aidr.today/5acdf59a?lang=en"
title: "Tree Navigation Without LLM Summaries: A Matched-Cost Study of Hierarchical Retrieval for Long-Document QA"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-07T04:00:00.000Z"
category: "Data"
topics: ["llm","data-engineering"]
source_urls: ["https://arxiv.org/abs/2610.06902"]
summary: "NavTree, a leaves-only hierarchical retriever, builds a deterministic balanced segment tree over document chunks with zero language-model calls at indexing and uses a hybrid lexical-and-dense frontier walk for navigation, emitting only leaf chunks to the reader. On a matched-cost evaluation, NavTree is the strongest matched-cost hierarchical retriever and ties the strongest flat baseline, demonstrating that summary trees' main benefit in long-document QA comes from navigation rather than generated summary content."
---

# Tree Navigation Without LLM Summaries: A Matched\-Cost Study of Hierarchical Retrieval for Long\-Document QA

> [Open the canonical story](<https://aidr.today/5acdf59a?lang=en>)

**Published:** 2026-10-07T04:00:00.000Z
**Category:** Data
**Topics:** llm, data\-engineering

## Summary

NavTree, a leaves\-only hierarchical retriever, builds a deterministic balanced segment tree over document chunks with zero language\-model calls at indexing and uses a hybrid lexical\-and\-dense frontier walk for navigation, emitting only leaf chunks to the reader\. On a matched\-cost evaluation, NavTree is the strongest matched\-cost hierarchical retriever and ties the strongest flat baseline, demonstrating that summary trees' main benefit in long\-document QA comes from navigation rather than generated summary content\.

## Sources

- [Story source](<https://arxiv.org/abs/2610.06902>)