Information for
Quick links
Faculties
How can AI systems – particularly Large Language Models (LLMs) – be modified to respond reliably and safely to specific requirements? And how exactly does this process work; how can it be made traceable, understandable, and reproducible? Marius Mosbach’s research addresses these questions surrounding the fine-tuning of LLMs – with the goal of turning the adaptation that is necessary for all applications into a principled science and use that understanding to build more reliable and adaptable AI systems.
After several years as postdoctoral researcher at Mila – the Quebec Artificial Intelligence Institute and postdoctoral fellow at McGill University in Montréal, Canada, the researcher is looking forward to returning to Saarland University, where he studied and earned his Ph.D.: “It’s wonderful, of course, that I was able to secure the professorship. I’m glad to be back in Saarland and to be able to contribute to its strong research landscape. I’m most looking forward to building my own group and working with students on exciting topics related to language models. The collaborative aspect of research is very important to me.”
To move closer to the long-term vision of AI systems that adapt transparently, robustly, and continuously over time, Marius Mosbach’s research group will work along three directions
1. Interpretability: How do LLMs actually work, and how does adaptation change them?
2. Generalization: How and when can models function reliably even outside their training data?
3. Continual learning: How can systems be built that do not “freeze” at the moment training ends, but instead autonomously absorb new knowledge during deployment, update or remove outdated facts, and improve from interaction?
Marius Mosbach has made a name for himself in academic circles through, amongst other things, LLM2Vec, a method that enables large language models to be transformed into powerful text encoders. Furthermore, his publications have been honored several times.
In addition to his professorship in the Department of Language Science and Technology, Marius Mosbach will also serve as scientific director at the German Research Center for Artificial Intelligence (DFKI). This means he will be closely connected in two ways to both the international research community and AI research at Saarland Informatics Campus (SIC).
We warmly welcome Marius Mosbach to the department and look forward to a productive collaboration!
More about Marius Mosbach:
Personal website
Official press release of Saarland University