Friedrich Schiller University Jena is a major German research university that hosts the LAMALab, a research group led by Dr. Kevin Jablonka focused on AI-accelerated materials discovery and LLM benchmarking in chemistry.
Friedrich Schiller University Jena is a major German research university that hosts the LAMALab, a research group led by Dr. Kevin Jablonka focused on AI-accelerated materials discovery and LLM benchmarking in chemistry.
People– no linked people
Updated 04/02/26Funding Details
Updated 04/02/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 04/02/26Friedrich Schiller University Jena (FSU Jena), founded in 1558, is one of ten oldest universities in Germany and a major public research institution in Thuringia. With approximately 16,500 students, 10,000 employees, and a total annual budget of EUR 572 million (2024), the university maintains research partnerships in over 73 countries and hosts significant third-party funded research exceeding EUR 183 million annually. The university's most directly AI safety-relevant work is conducted by the LAMALab (Laboratory for AI in Materials Science), an independent research group led by Dr. Kevin Maik Jablonka. Jablonka holds a tenure-track position as group leader at HIPOLE Jena (Helmholtz Institute for Polymers in Energy Applications), affiliated with FSU Jena, since June 2023. The group currently has 11 members including 9 PhD students and 1 postdoctoral researcher. The LAMALab's primary relevance to AI safety stems from its systematic work benchmarking large language models on consequential scientific tasks. In December 2024, Open Philanthropy awarded a grant of approximately €786,060 over two years to FSU Jena for Dr. Jablonka's project developing an analytical chemistry benchmark to assess whether LLM agents can solve organic structural analysis problems. This was funded as part of Open Philanthropy's focus on potential risks from advanced AI, via a request for proposals for projects benchmarking LLM agents on real-world tasks. The group has published ChemBench in Nature Chemistry (2025), a framework with over 2,700 curated question-answer pairs across chemistry disciplines, finding that top LLMs outperform expert human chemists on average. They also developed MaCBench, a benchmark for multimodal vision-language model evaluation in chemistry. Dr. Jablonka has secured over EUR 4 million as principal investigator from sources including the Carl-Zeiss Foundation, Open Philanthropy, Intel-Merck, and the Helmholtz Foundation Model Initiative. He also serves as an independent red teamer for OpenAI since July 2024. At the institutional level, FSU Jena joined ELSA (European Lighthouse on Secure and Safe AI) in 2025, a European Union network of excellence focused on trustworthy and secure AI. The university co-leads the ELLIS Unit Jena with the Max Planck Institute for Biogeochemistry and the German Aerospace Center, focusing on AI for environmental and climate science applications.
Theory of Change
Updated 04/02/26By developing rigorous, expert-level benchmarks for LLM capabilities in high-stakes scientific domains (chemistry), the LAMALab produces tools that allow the AI safety and research community to accurately assess where LLMs are capable versus overconfident, and to track capability thresholds over time. Reliable benchmarks in consequential domains help identify when AI systems may be deployed prematurely or dangerously, and support informed governance and safety decisions. Dr. Jablonka's red-teaming work at OpenAI extends this capacity-assessment approach directly into frontier model development.
Grants Received
Updated 04/02/26Projects– no linked projects
Updated 04/02/26Discussion
No comments yet. Be the first to share your thoughts.