PhD/Postdoc Positions on LLM Theory, Interpretability, and Safety at Saarland University

[Michael H]

The Language, Computation, and Cognition Lab (LaCoCo) at Saarland Informatics Campus (Saarbrücken, Germany), directed by Michael Hahn, invites applications for fully funded PhD and postdoc positions, with a flexible start date.


RESEARCH AREAS

We’re especially excited about projects on:

• Mathematical foundations of large language models (LLMs), e.g., computational complexity, learning theory
• New architectures and efficient reasoning for LLMs
• Mechanistic interpretability
• Theoretical foundations for AI safety

You’ll have substantial freedom to shape your topic within the lab’s scope.

Some representative publications from the past 3 years:

• Discovering Interpretable Algorithms by Decompiling Transformers to RASP (ICML 2026, https://arxiv.org/abs/2602.08857)
• Why are Sensitive Functions Hard for Transformers? (ACL 2024, https://aclanthology.org/2024.acl-long.800/)
• InversionView: A General-Purpose Method for Reading Information from Neural Activations (NeurIPS 2024, https://openreview.net/forum?id=clDGHpx2la)
• Contextualize-then-Aggregate: Circuits for In-Context Learning in Gemma-2 2B (COLM 2025, https://openreview.net/forum?id=rGNAyHReSg)
• On the Ability of Transformers to Verify Plans (ICML 2026, https://arxiv.org/pdf/2603.19954)

ABOUT THE LAB

Our research has been awarded an Emmy Noether grant (DFG), the Heinz Maier-Leibnitz Preis (DFG), and paper awards at ACL 2024 and Mechanistic Interpretability Workshop 2024. We regularly publish at top venues (e.g., 11 NeurIPS/ICML/ICLR papers over the past year).

Learn more about the lab at https://lacoco-lab.github.io/

HOW TO APPLY

Please follow the instructions here: https://lacoco-lab.github.io/joining/

For questions, please contact Michael Hahn at mh...@lst.uni-saarland.de

-- 

Michael Hahn

Tenure-Track Professor

Saarland Informatics Campus

Saarland University

Group: https://lacoco-lab.github.io/

Personal: https://www.mhahn.info/