Follow the Entities: A Corpus Map for Agentic Search

2026-09-29

AI Summary

This study addresses the challenges faced by agents in linking cross-document evidence within flat document corpora, including redundant relation discovery and excessive token consumption. To overcome these limitations, this work proposes CorpusMap, a navigation layer that constructs a corpus graph anchored on entities. By integrating entity disambiguation with offline indexing techniques, CorpusMap shifts relation discovery from inference time to an offline precomputation stage, thereby enabling link reuse across queries. Experimental evaluations across seven models and three benchmarks demonstrate that CorpusMap significantly improves both evidence retrieval efficiency and answer quality while substantially reducing average token consumption. Furthermore, it consistently outperforms existing navigation approaches.

Abstract

Answering questions and completing tasks over large document collections often requires connecting evidence spread across multiple documents, such as a project's approval recorded in one, its requirements in another, and its latest status in a third. Recent LLM agents approach this by iteratively searching the full corpus rather than reading only a fixed set of top-ranked documents. However, when the corpus is exposed only as a flat collection of files, a relevant document gives no indication of how it relates to others, so the agent must rediscover these relationships for every query, often missing complementary evidence while simultaneously consuming substantial additional tokens. To address this, we introduce CorpusMap, a navigation layer that organizes the corpus around its recurring entities, which are identifiable from the documents themselves and can link a single document to many others across sources. Specifically, CorpusMap represents each recurring entity as an Entity Page that aggregates information about it and links to every document that refers to it, forming a graph between entities and documents that the agent can traverse to gather otherwise disconnected evidence. Moreover, since CorpusMap is constructed offline by resolving mentions of the same entity across documents, its links are shared across queries rather than rediscovered repeatedly at inference time. Using 7 different models with 3 benchmark datasets, we show that CorpusMap improves both evidence discovery and answer quality over raw-corpus agentic search while using fewer tokens on average, and further outperforms 4 alternative navigation layers, suggesting that entities serve as effective anchors for navigating large document collections.

Problem

• Agentic Search
• Multi-document Reasoning
• Corpus Navigation
• Evidence Discovery
• Large Language Model Agents

Innovation

• CorpusMap
• Agentic Search
• Entity Pages
• Navigation Layer
• Evidence Discovery