If you missed the seminar with Helge Stein on March 24, 2025, you can now watch the recording online. His lecture, "The Engineering of Research by Horizontal and Vertical Integration", is available on our YouTube channel.

We are excited to share that a new FAIRmat paper has just been published in Nature Chemistry!

The paper, titled "A framework for evaluating the chemical knowledge and reasoning abilities of large language models against the expertise of chemists", introduces ChemBench — a comprehensive benchmarking framework designed to assess the performance of large language models (LLMs) compared to trained chemists across a wide range of chemistry tasks.

Led by Kevin Maik Jablonka, the FAIRmat AI Toolkit task leader, the team of researchers addresses the critical need for systematic evaluation tools to understand the strengths and limitations of LLMs in the chemical sciences. In the absence of any existing standardized benchmarks tailored to chemistry, ChemBench fills a significant gap.

To develop ChemBench, the authors curated over 2,700 question–answer pairs covering a variety of chemistry topics and skill levels. They developed a web-based platform that can render molecules correctly, track evaluation progress, and host a public leaderboard, all while maintaining data privacy.

Crucially, the authors compared the performance of leading LLMs, both open-source and closed-source, against a group of 19 expert chemists, some of whom were permitted to use tools such as web searches to simulate realistic workflows. The results are striking:

"On average, the best models outperformed the best human chemists in this study — yet they also struggled with some basic tasks and often gave overconfident predictions."

Performance of models and humans on ChemBench-Mini

These insights highlight the impressive potential of AI in chemistry, as well as its current limitations. They also emphasize the importance of ongoing development and deployment of AI in chemistry, as well as potential changes in chemistry education and AI integration strategies.

• Behind the ChemBench Paper: Measuring AI's Chemistry Capabilities
• Explore ChemBench 

To bring the power of the NOMAD ecosystem directly to researchers, we set off on the NOMAD Tour from May 19 to 22, with stops in Mainz, Darmstadt, and Heidelberg. Led by Hampus Näsström and Joseph F. Rudzinski, each event welcomed local research communities for an engaging introduction to the capabilities of NOMAD.

Each stop featured two parts. The first was an overview talk with an interactive Q&A, tracing NOMAD’s evolution from a repository for ab initio calculations to a robust, FAIR-compliant research data management ecosystem. The talks highlighted how NOMAD helps researchers streamline their workflows, ensure data quality, and meet modern open science standards.

The second part was a hands-on tutorial, allowing participants to explore NOMAD’s tools in depth. Attendees selected from four key focus areas: project and workflow management, Electronic Lab Notebooks (ELNs), plugin development, and local infrastructure setup. These sessions were tailored to both newcomers and experienced users, offering practical guidance and one-on-one support.

With enthusiastic participation in every city, the tour confirmed the growing interest in digital research infrastructure. We thank everyone who joined us on the road – and if you missed out, stay tuned. NOMAD may be coming to your city next!

The presentation slides are already available for download. 

In our podcast series Pioneers in electronic-structure theory, FAIRmat host Miguel Marques speaks with leading figures who played a key role in developing the methods and codes that define electronic-structure theory today. This work includes pioneering density-functional theory and methods beyond to treat many-body effects in ground-state and excitations.

In the fourth episode, Miguel Marques interviews Giovanni Vignale, visiting professor at the National University of Singapore. Vignale shares personal insights from his student days, including his move from Italy to the USA for his PhD. He also reflects on how early-career research has changed since the 1980s. In the second part of the episode, Vignale discusses his writing beyond academic journals and his passion for sports.

Find the podcast on YouTube or enjoy the audio-only version on Spotify!

Last week, FAIRmat hosted the NFDI Physical Sciences joint colloquium with Prof. Isao Tanaka, who gave a lecture titled "Recommender System for Discovery of New Inorganic Compounds." If you missed it, you can now watch the recording on our YouTube channel!

Prof. Claudia Draxl, FAIRmat’s spokesperson and Einstein Professor of Physics at Humboldt-Universität zu Berlin, has been elected to the German National Academy of Sciences Leopoldina, one of the highest scientific honors in Germany.

This prestigious recognition highlights her outstanding scientific contributions to condensed-matter theory, computational materials science, and the advancement of data-centric approaches to materials research.

In her role at FAIRmat, Prof. Draxl has actively shaped the development of research data infrastructure for materials science. Under her leadership, FAIRmat supports researchers in making their data Findable, Accessible, Interoperable, and Reusable (FAIR), promoting transparency, reproducibility, and reusability in scientific research.

Her election to the Leopoldina reflects not only her scientific excellence, but also her pioneering efforts to build a sustainable and open-data culture in science. It is a proud moment for the entire FAIRmat community.

We warmly congratulate Prof. Draxl on this well-deserved honor!

Recordings of FAIRmat Tutorial 16 are now available on the FAIRmat and NOMAD YouTube channel! The full playlist includes:

• Introduction to NOMAD public service and its main functionalities by Adrianna Wojas
• Exploring NOMAD entries with search filters and interactive widgets by Siamak Nakhaie
• From raw data files to published datasets by Ahmed Mansour
• Guiding through the steps for documenting research using NOMAD ELN functionality by Siamak Nakhaie

Follow along with the tutorial using the full tutorial guide. Slides and example files are available here.

For more FAIRmat hands-on tutorials, visit our website.

The FAIRmat team took part in the DPG Spring Meeting of the Condensed Matter Section at the University of Regensburg from March 16 to 21, 2025. Our colleagues from different Areas delivered multiple talks on recent developments in NOMAD to enable FAIR research data management in computational materials science and experimental research. Their presentations provided valuable insights into the practical implementation of the FAIR principles in various aspects of materials science.

As in previous years, together with DAPHNE4NFDI, we hosted a joint information booth at the DPG exhibition for scientific instruments and literature. The strong interest in NOMAD and research data management from the community keeps inspiring us to push forward!

A pleasant learning environment highlighted our tutorial session on Using NOMAD’s Workflow Utilities to Improve Data Management and Facilitate Discovery in Materials Science. If you missed our DPG Tutorial, it is available online for self-learning.

Last but not least, two insightful symposia stood out at this year's DPG Spring Meeting. The first focused on AI-driven Materials Design and was organized by Jörg Neugebauer, Silvana Botti, and Luca Ghiringhelli. The second, organized by Frank Schreiber, Bridget Murphy, and Claudia Draxl in collaboration with DAPHNE4NFDI, explored Pushing the Boundaries of FAIR Data Practices for Condensed Matter Insights.

We thank the organizers for a fantastic event and look forward to the next meeting in 2026!

We are excited to share the recordings of users' talks and invited speakers' presentations from our latest Users Meeting last November! The full playlist includes:

• FAIR Research Data Management with FAIRmat and NOMAD – Ahmed Mansour
• The FAIRification of PV research data – Eva Unger
• Enabling high-throughput materials discovery of phosphosulfides by developments in FAIR data management, visualization and analysis in NOMAD – Lena Mittmann
• Martignac: Computational workflows for reproducible, traceable, and composable coarse-grained Martini simulations – Tristan Bereau
• Achieving semantic interoperability in materials science data and simulation workflows – Abril Azocar Guzman
• An overview of NFDI4Cat´4 services and tools with a special focus on Voc4Cat the shared vocabularies for catalysis and related disciplines – David Linke

Subscribe to the FAIRmat and NOMAD YouTube channel now to stay updated on our latest insightful videos!

Last week, the FAIRmat team participated in Love Data Week 2025, an international campaign promoting research data and effective data management.

Siamak Nakhaie and Ahmed Mansour hosted insightful coffee talks on researchers' rights to the data they create and the challenges of managing diverse, large datasets. Both talks are available on our YouTube channel.

Meanwhile, our colleague Julia Schumann participated as an invited speaker at the NFDI4Cat consortium workshop, exploring various RDM solutions for catalysis research.

Let’s keep the spirit of Love Data Week alive and celebrate our passion for data all year round!