HERMES: a multi-agent framework for structured knowledge extraction from ultra-long documents in geoscience

# HERMES: a multi-agent framework for structured knowledge extraction from ultra-long documents in geoscience
Source: [https://arxiv.org/abs/2608.14055](https://arxiv.org/abs/2608.14055)
Authors:[Ziqi Song](https://arxiv.org/search/cs?searchtype=author&query=Song,+Z),[Zongyuan Xiang](https://arxiv.org/search/cs?searchtype=author&query=Xiang,+Z),[James G\. Ogg](https://arxiv.org/search/cs?searchtype=author&query=Ogg,+J+G),[Bruce S\. Lieberman](https://arxiv.org/search/cs?searchtype=author&query=Lieberman,+B+S),[Gabi Ogg](https://arxiv.org/search/cs?searchtype=author&query=Ogg,+G),[Natalia López Carranza](https://arxiv.org/search/cs?searchtype=author&query=Carranza,+N+L),[Wen Du](https://arxiv.org/search/cs?searchtype=author&query=Du,+W),[Yufei Ye](https://arxiv.org/search/cs?searchtype=author&query=Ye,+Y),[Shuan Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+S),[Zhong Peng](https://arxiv.org/search/cs?searchtype=author&query=Peng,+Z),[Shaoqi Yu](https://arxiv.org/search/cs?searchtype=author&query=Yu,+S),[Juye Wei](https://arxiv.org/search/cs?searchtype=author&query=Wei,+J),[Ying Zhou](https://arxiv.org/search/cs?searchtype=author&query=Zhou,+Y),[Jieping Ye](https://arxiv.org/search/cs?searchtype=author&query=Ye,+J),[Jiang Yang](https://arxiv.org/search/cs?searchtype=author&query=Yang,+J)

[View PDF](https://arxiv.org/pdf/2608.14055)

> Abstract:Authoritative scientific knowledge in geoscience remains largely trapped in legacy monographs and historical literature, where unstructured text and complex layouts hinder computational access\. We introduce HERMES, a scalable multi\-agent framework that extracts structured data from ultra\-long scientific documents\. Using a coordinating large language model, HERMES integrates domain constraints, validation rules and evidence tracing within a unified document\-level extraction process that incorporates parsed text, tables, figures and captions\. Applied to the 55\-volume Treatise on Invertebrate Paleontology, the system produced a structured database of 32,277 fossil taxonomic entities and 451,878 attributes, released online at[this https URL](https://treatise.geolex.org/)\. Extraction performance remained stable across fossil groups \(average F1 scores of approximately 0\.90 for entities and 0\.91 for attributes\), improving per\-volume efficiency approximately sixfold relative to the tested fully manual baseline\. Evaluation in palaeomagnetism and geochemistry, conducted without additional model training, demonstrated transfer across distinct geoscience domains\. This work provides a practical pathway to transform historical scientific literature into FAIR\-oriented structured data, offering a sustainable infrastructure for data\-intensive disciplines and large\-scale knowledge integration\.

## Submission history

From: Zongyuan Xiang \[[view email](https://arxiv.org/show-email/1754d96a/2608.14055)\] **\[v1\]**Fri, 14 Aug 2026 07:59:58 UTC \(3,857 KB\)