We are looking for an eager Master student to join us for three to six months in the period of January to October 2026 for a research internship and to work on cutting edge methods for uncertainty estimation in Large Language Models. Specifically, vanilla LLMs do not have a way of distinguishing between knowledge and data uncertainty, which has been shown to be particularly useful in downstream tasks. Methods that estimate these types of uncertainty rely on computationally expensive ensemble models that are impractical for LLMs. We would like the student to contribute to our ongoing work on using Dirichlet estimation for approximation to tackle this problem. An ideal candidate would be close to finishing their master’s degree in Computer Science, Mathematics or Engineering, have prior experience working with LLMs and neural networks as well as Bayesian methods of estimation.

The successful master student will receive a payment of 1000-1200 € per month depending on the prior experience of the selected candidate. Depending on the travel distance of the successful candidate a lump sum of 100-450 € will be given to cover travel costs.

Please send a motivation letter, CV, and transcripts from all studies so far by 17^th of November 2025 to Prof. Dr. Milica Gasic. If necessary, candidates will be interviewed between 19^th and 20^th of November 2025. The decision will be made on 21^st of November 2025. For any questions, please contact Prof. Dr. Milica Gasic directly.