Academic research assistant (m/f/x)

Saarland University is a campus university with an international reputation for research excellence, particularly in computer science and in the life sciences and nanosciences. The university is also distinguished by its close ties to France and its strong European focus. Around 17,000 students, studying over one hundred different academic disciplines, are currently enrolled at Saarland University. Saarland University is officially recognized as one of Germany’s family-friendly higher-education institutions and with a combined workforce of more than 4000 it is one of the largest employers in the region.

The Department of Computer Science at Saarland University is inviting applications for the following position commencing 01 January 2027 Academic research assistant (m/f/x) Reference number W2936, salary in accordance with the German TV-L salary scale1, pay grade: E-13 TV-L; for 3 years, contract renewable until 12/2030, employment: 100 % of standard working time.

Workplace/Department:

The Chair of Computer Science and Computational Linguistics invites excellent students holding master’s degree for for a fully funded one doctoral position in the field of large language models, natural language processing, computational pragmatics, and knowledge representation.

Prof. Vera Demberg’s lab covers the highly interdisciplinary and fast-paced fields of psycholinguistics and computational linguistics.Saarland University is one of the leading centres for computational linguistics and computer science in Europe and offers a dynamic and stimulating research environment. It is famous for its interdisciplinary research in linguistics, computation and cognition.

The lab is affiliated with both the Department of Computer Science and with the Department of Language Science and Technology. The Department of Language Science and Technology organizes about 100 research staff in ten research groups in the fields of computational linguistics, psycholinguistics, speech processing, and corpus linguistics.

Both departments are part of the Saarland Informatics Campus, which brings together 800 researchers and 2000 students from 81 countries. We collaborate closely with the university's Department of Computer Science, the Max 1 TV-L = collective agreement on remuneration of public sector employees in the German Länder Planck Institute for Informatics, the Max Planck Institute for Software Systems, and the German Research Center for Artificial Intelligence (DFKI).

Our researchers and students come from all over the world, and our primary working language is English.

The focus of the project: Cooperative User Interfaces as Rational Agents

As large language model–based agents increasingly act autonomously and adapt to their users, a central challenge for cooperative human–AI interaction is enabling systems to reason explicitly about what a given user knows and believes, and to tailor their communication accordingly. Our project addresses this challenge by grounding the behaviour of cooperative LLM agents in personal knowledge bases (PKBs): explicit, inspectable representations of a user's knowledge, beliefs, and expertise. This PhD position should develop the reasoning and generation methods that allow an agent to exploit such representations to communicate cooperatively, and will investigate how faithfully LLMs actually use this knowledge.

The successful candidate will address four related research challenges. The first concerns the empirical characterisation of knowledge-dependent inference: building on natural language inference datasets, the goal is to investigate how a comprehender's background knowledge shapes the inferences they draw. The second challenge concerns the computational modelling of these differences. Here the doctoral researcher will develop knowledge-graph–augmented LLMs that retrieve information about a user's knowledge from their PKB, and predict how a given text would be interpreted under different knowledge states. The third step aims to turn prediction into cooperative generation. Drawing on the Rational Speech Acts (RSA) framework, the work will identify gaps in a user's knowledge relevant to a target message, determine what background information must be made explicit to prevent misinterpretation, and iteratively refine generated utterances by simulating how they would be understood by the modelled user.

Finally, the position engages with the faithfulness and controllability of these systems through mechanistic interpretability: systematically assessing when LLMs rely on PKB-retrieved information versus their parametric knowledge, localising the responsible circuits, and exploring activation-steering methods to make PKB-grounded behaviour reliable and actionable.

Job requirements and responsibilities:

The position is intended to start as soon as the beginning of January 2027 (start-date is negotiable, i.e., a later start is possible).

Prior experience of research activities in large language models, natural language processing, computational pragmatics, and knowledge representation. Experience with RSA-style modelling, knowledge graphs, or mechanistic interpretability is advantageous but not required.

Prior experience in authoring scientific papers at a top venue is a plus.

Strong grades from a top degree program in a relevant field.

The research will be conducted in English, and a successful applicant must have excellent spoken and written proficiency in English.

Strong motivation for scientific research.

Strong interest in the research topics studied by the group.

Your academic qualifications:

Completed university studies in : A master's degree (holding or nearing completion of) in computational linguistics, computer science, or a cognate discipline, with strong programming skills and a demonstrated interest in one or more of the following: large language models, natural language processing, computational pragmatics, and knowledge representation. Experience with RSA-style modelling, knowledge graphs, or mechanistic interpretability is advantageous but not required.

The successful candidate will also be expected to:

Contribute to activities of the research group in the areas of research, teaching and mentoring

Applicants are requested to submit their application, including: – a cover letter that specifies why you would like to work on this topic and what qualifies you for it, – an academic CV, – a list of academic publications, – your MSc thesis (or a current draft), – copies of academic degree certificates, – names of two potential references. Please also include a 2-page research proposal in your application which outlines how you would approach the topic.

What we can offer you:

Access to state-of-the art research and training facilities and a generous conference and travel budget

Immersion in an interdisciplinary community made up of cognitive and computational researchers

A flexible work schedule allowing you to balance work and family, among other things the possibility of teleworking

Secure and future-oriented employment with attractive conditions

A broad range of further education and professional development programmes (for example language courses)

An occupational health management model with numerous attractive options, such as our university sports programme

Supplementary pension scheme (RZVK)

Discounted tickets on local public transport services (‘Job-Ticket‘ of the saarVV)

Job bike leasing (JobRad)

We look forward to receiving your meaningful online application (in a PDF file) by 31.10.2026 to [email protected]. Please include the reference number W2936 in the subject line of the e-mail.

*If you have any questions, please contact Prof. Vera Demberg for assistance.*

*Pay grade classification is based on the particular details of the position held and the extent to which the applicant meets the requirements of the pay grade within the TV-L salary s