Marius Mosbach to join Saarland University and DFKI, seeking PhD students for LLM adaptation and interpretability research

Marius is a fantastic advisor, collaborator, colleague, and friend. Bittersweet to see him leave Montreal, but excited for all the amazing things he'll do at Saarland University and DFKI 🚀. He's hiring PhD students in LLM adaptation, interpretability, and continual learning. Apply now!

@mariusmosbach @DFKI congratulations!!! 🎉 we should collaborate again ;)

🚨 I’m happy to share that I'm joining Saarland University as a tenure-track Professor in the Department of Language Science and Technology, and the German Research Center for AI (@DFKI) as a Scientific Director, starting October 2026! 🚨 1/6

I'm looking for a motivated PhD student to start in October 2026, plus research interns, with additional PhD funding coming next year. If you're excited about model adaptation, interpretability, or continual learning, please fill out this form 👉 https://forms.gle/4r8uvX3uqoPGmKMu7 5/6

My group will work on building more reliable and adaptable AI systems, especially LLMs. Our main focus is the science of model adaptation 🤖🧬: understanding how models generalize and become specialized, capable, and safe. More on our research 👉 https://mariusmosbach.com/group 2/6

A second focus of the group is the interpretability of LLMs 🔎. In particular, making interpretability research more actionable. We want insights into how models work to lead to concrete improvements: better ways to adapt, steer, and trust them. 3/6

I'm also expanding into continual learning 🔄: the current adaptation pipeline is sequential; models are frozen once training ends. We want systems that keep learning after deployment, absorbing new knowledge and updating outdated facts without expensive retraining. 4/6

@mariusmosbach @DFKI Gratuliere!! 👏

You can also find me at #ICML2026 in Seoul 🇰🇷 (Jul 6–11) presenting 3 papers:

📍 LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs

📍 Operationalising the Superficial Alignment Hypothesis via Task Complexity

📍 Interpretability Can Be Actionable. 6/6

@mariusmosbach @DFKI congratulations likely to be professor! 🍻

@mariusmosbach @DFKI 🍻

@mariusmosbach @DFKI Congrats Marius!

@mariusmosbach @DFKI woohoo congrats!!!

@mariusmosbach @DFKI wow, congrats!

@mariusmosbach @DFKI Amazing, congratulations! 🥳

@mariusmosbach @DFKI ngl Marius's gonna be the best prof you can find!

@mariusmosbach @DFKI Congrats!

@mariusmosbach @DFKI So happy for you, Marius 🥹🎉

@mariusmosbach @DFKI Woohoo!! 🎉 Congrats Marius!

@mariusmosbach @DFKI Congrats!!