ChemPile: A 250GB Diverse and Curated Dataset for Chemical Foundation Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mirza, Adrian, Alampara, Nawaf, Ríos-García, Martiño, Abdelalim, Mohamed, Butler, Jack, Connolly, Bethany, Dogan, Tunca, Nezhurina, Marianna, Şen, Bünyamin, Tirunagari, Santosh, Worrall, Mark, Young, Adamo, Schwaller, Philippe, Pieler, Michael, Jablonka, Kevin Maik
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910951697022976
author Mirza, Adrian
Alampara, Nawaf
Ríos-García, Martiño
Abdelalim, Mohamed
Butler, Jack
Connolly, Bethany
Dogan, Tunca
Nezhurina, Marianna
Şen, Bünyamin
Tirunagari, Santosh
Worrall, Mark
Young, Adamo
Schwaller, Philippe
Pieler, Michael
Jablonka, Kevin Maik
author_facet Mirza, Adrian
Alampara, Nawaf
Ríos-García, Martiño
Abdelalim, Mohamed
Butler, Jack
Connolly, Bethany
Dogan, Tunca
Nezhurina, Marianna
Şen, Bünyamin
Tirunagari, Santosh
Worrall, Mark
Young, Adamo
Schwaller, Philippe
Pieler, Michael
Jablonka, Kevin Maik
contents Foundation models have shown remarkable success across scientific domains, yet their impact in chemistry remains limited due to the absence of diverse, large-scale, high-quality datasets that reflect the field's multifaceted nature. We present the ChemPile, an open dataset containing over 75 billion tokens of curated chemical data, specifically built for training and evaluating general-purpose models in the chemical sciences. The dataset mirrors the human learning journey through chemistry -- from educational foundations to specialized expertise -- spanning multiple modalities and content types including structured data in diverse chemical representations (SMILES, SELFIES, IUPAC names, InChI, molecular renderings), scientific and educational text, executable code, and chemical images. ChemPile integrates foundational knowledge (textbooks, lecture notes), specialized expertise (scientific articles and language-interfaced data), visual understanding (molecular structures, diagrams), and advanced reasoning (problem-solving traces and code) -- mirroring how human chemists develop expertise through diverse learning materials and experiences. Constructed through hundreds of hours of expert curation, the ChemPile captures both foundational concepts and domain-specific complexity. We provide standardized training, validation, and test splits, enabling robust benchmarking. ChemPile is openly released via HuggingFace with a consistent API, permissive license, and detailed documentation. We hope the ChemPile will serve as a catalyst for chemical AI, enabling the development of the next generation of chemical foundation models.
format Preprint
id arxiv_https___arxiv_org_abs_2505_12534
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ChemPile: A 250GB Diverse and Curated Dataset for Chemical Foundation Models
Mirza, Adrian
Alampara, Nawaf
Ríos-García, Martiño
Abdelalim, Mohamed
Butler, Jack
Connolly, Bethany
Dogan, Tunca
Nezhurina, Marianna
Şen, Bünyamin
Tirunagari, Santosh
Worrall, Mark
Young, Adamo
Schwaller, Philippe
Pieler, Michael
Jablonka, Kevin Maik
Machine Learning
Foundation models have shown remarkable success across scientific domains, yet their impact in chemistry remains limited due to the absence of diverse, large-scale, high-quality datasets that reflect the field's multifaceted nature. We present the ChemPile, an open dataset containing over 75 billion tokens of curated chemical data, specifically built for training and evaluating general-purpose models in the chemical sciences. The dataset mirrors the human learning journey through chemistry -- from educational foundations to specialized expertise -- spanning multiple modalities and content types including structured data in diverse chemical representations (SMILES, SELFIES, IUPAC names, InChI, molecular renderings), scientific and educational text, executable code, and chemical images. ChemPile integrates foundational knowledge (textbooks, lecture notes), specialized expertise (scientific articles and language-interfaced data), visual understanding (molecular structures, diagrams), and advanced reasoning (problem-solving traces and code) -- mirroring how human chemists develop expertise through diverse learning materials and experiences. Constructed through hundreds of hours of expert curation, the ChemPile captures both foundational concepts and domain-specific complexity. We provide standardized training, validation, and test splits, enabling robust benchmarking. ChemPile is openly released via HuggingFace with a consistent API, permissive license, and detailed documentation. We hope the ChemPile will serve as a catalyst for chemical AI, enabling the development of the next generation of chemical foundation models.
title ChemPile: A 250GB Diverse and Curated Dataset for Chemical Foundation Models
topic Machine Learning
url https://arxiv.org/abs/2505.12534