---
title: "STAIR Beats Every RAG Baseline by Reading Documents Like a Table of Contents"
subtitle: "A new generative retriever leverages the table of contents to hit 82.6% Recall@1 on a fresh 18-book benchmark, with hallucinations under 0.05%."
url: "https://alphasignal.ai/news/stair-beats-every-rag-baseline-by-reading-documents-like-a-table-of-contents"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-09-03T14:02:53.000Z"
updated: "2026-09-10T00:06:09.095Z"
type: "Paper"
topics:
  - "Retrieval"
  - "Benchmarks"
subtopics:
  - "Rag"
  - "Document Understanding"
  - "Embeddings"
upvotes: 161
source_url: "https://arxiv.org/abs/2609.03874"
comments: 0
truncated: true
---

# STAIR Beats Every RAG Baseline by Reading Documents Like a Table of Contents

> A new generative retriever leverages the table of contents to hit 82.6% Recall@1 on a fresh 18-book benchmark, with hallucinations under 0.05%.

Most retrieval augmented generation pipelines treat a document like a stack of Lego bricks. They slice it into fixed-length chunks, embed each one, and hope cosine similarity does the rest. In the process, the map of the document, its chapters, sections, and subsections, gets thrown out. A new paper argues that map is exactly what a retriever should be reading.

The system is called [STAIR (STructure Aware Information Retriever)](https://arxiv.org/abs/2609.03874), and it comes paired with a benchmark called SearchTome. STAIR hits Recall@1 of 82.6% on SearchTome versus 76.9% for DSI, and beats other strong baselines including BM25 (59.5%), DPR (68.7%), and out-of-the-box Mistral (13.8%). The gap over DSI is reported as statistically significant, and the hallucination rate sits below 0.05 percent.

## Chunking throws away the map

LLMs keep getting better at long context, but they still suffer from the "lost in the middle" problem, where facts buried deep in a huge prompt get ignored. Precise retrieval remains a hard requirement. The standard recipe of splitting a book into 512-token windows discards the semantic hierarchy that the author put there on purpose.

Think about how a human answers a question from a textbook. You do not fuzzy-match paragraphs. You skim the table of contents, jump to the likely chapter, then narrow down. Length-based chunking flattens that hierarchy into a bag of passages. STAIR instead lets an LLM exploit the global structure, such as a Table of Contents, to store and retrieve information from its own weights.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [STAIR: A novel dataset and LLM based retriever for document structure augmentation](https://arxiv.org/abs/2609.03874)
- [STAIR (HTML version)](https://arxiv.org/html/2609.03874v1)
- [STAIR on OpenReview](https://openreview.net/forum?id=reuiQzjULZ)