Seminar Assignment Summer 2026
The central registration for all computer science seminars will open on Feb 10th.
This system is used to distribute students among the available seminars offered by the CS department. To register for any of the seminars, you have to register here until April 7th, 23:55 CET. You can select which seminar you would like to take, and will then be automatically assigned to one of them by April 9th.
Please note the following:
We aim to provide a fair mapping that respects your wishes, but at the same time also respects the preferences of your fellow students. Experience has shown that particular seminars are more popular than others, yet these seminars cannot fit all students. Please only select seminars if you are certain that you actually do want to complete a seminar this semester. If you have already obtained sufficient seminar credits, or plan to take other courses this semester, please do not choose any seminars. Students who drop out of seminars take away places from those, who might urgently need a space or are strongly interested in the topic. We encourage those students who wish to take a seminar this semester, to select their preferences for all available seminars, which eases the process to assign students that do not fit the overly popular seminars to another, less crowded one. So if you are serious about taking a seminar this semester, please select at least three seminars (with priority from "High" to "Low"). If you urgently need to be assigned to a seminar in the upcoming semester, choose at least five seminars (with priority from "High" to "Low"). The system will then prioritize you for assigning a seminar (yet not necessarily your top choice). If you are really dedicated to one particular seminar, and you do not want any other seminar, please select the "No seminar" as second and third positive option. However, this may ultimately lead to the situation that you are not assigned to any seminar. Also, choosing "No seminar" as second/third option does not increase your chances of getting your first choice. The assignment will be performed by a constraint solver on April 9th, 2026. You will be added to the respective seminars automatically and be notified about this shortly thereafter. Please note that the assignment cannot be optimal for all students if you drop the assigned seminar, i.e., make only serious choices to avoid penalty to others.
Seminars
Advanced topics in causality and causal ML research by Sara Magliacane
In data science and real-world machine learning, there are many issues that are often neglected in standard machine learning courses. Many tasks are inherently trying to answer causal questions and gather actionable insights, even when there is not enough data to draw causal conclusions. Moreover, integrating causal reasoning in state-of-the-art AI can help improve the robustness and generalizability of current approaches, as well as imbue them with the strong theoretical guarantees typical of causality research.
In this seminar we will focus on the intersection of causality and machine learning research, by reading recent publications, discussing them and implementing extensions of current approaches in small groups.
In particular, we will investigate cutting-edge research in three different research directions:
1. Causal representation learning, i.e., the task of learning well-defined variables from high-dimensional data with theoretical guarantees on their disentanglement. An example would be learning the objects and their attributes from videos and actions of an agent interacting with a simulated kitchen environment. This is a very recent and promising research direction with many approaches focusing on different settings, including when we cannot perform actions, but there are still many open questions on how to develop practical methods that can scale to real-world settings.
2. Causal discovery, i.e., the task of learning causal relations and causal graphs from data, including the challenging setting in which we cannot do experiments. For example, a strong correlation between two variables X and Y is not enough to decide a policy in which we change X and expect to see an increase in Y (i.e. “correlation is not causation”). On the other hand, if we measure another variable Z that we know causes X, but does not have an effect on Y (i.e. an instrumental variable), we can discover under certain assumptions that X is the cause of Y, even if we haven’t performed any experiment. Current research in causal discovery extends this case to multiple variables, latent confounding and multiple observational and experimental datasets. While this is a more established research direction, there are still many open questions on the accuracy of current methods and their computational scalability, so there is still a lot of exciting research that can be done.
3. Downstream tasks for causal reasoning, e.g., transfer learning and compositional generalization based on causal representations, especially in reinforcement learning, or explainable AI with theoretical guarantees. While current research in causality in ML focuses mostly on developing new methods, there are still a lot of exciting opportunities on how research in causality can improve the robustness, generalization and safety of state-of-the-art AI methods.
Requirements: This seminary requires an excellent knowledge in machine learning, including having taken courses in core ML, elements of ML and linear algebra. A good knowledge of probability and statistics is also preferred.
The final evaluation will be based on in-class participation and a group project presentation.
Places: 12
Advanced Topics in Rational Intelligence: Prediction, Causation, Decision, Incentive, and Regulation All at Once by Krikamol Muandet
This seminar will explore topics central to the Rational Intelligence (RI) Lab (https://ri-lab.org/), spanning prediction, causation, and decision-making, as well as the incentive-aware learning and regulation surrounding algorithmic models.
In this seminar, students will develop core research skills by critically reading and analyzing research papers on selected topics, preparing and delivering presentations, and engaging in in-depth discussions with fellow participants and RI lab members.
Requirements: Students should possess sufficient knowledge of the fundamental concepts in machine learning.
Places: 15
AI Coding Assistants: Practices, Assumptions, and Implications by Sven Apel
Software development is inherently collaborative, relying on a range of practices and tools to support productivity, code quality, and knowledge sharing. In recent years, AI coding assistants such as GitHub Copilot have been integrated into everyday development workflows and are increasingly positioned as active collaborators rather than passive tools. Their growing use raises fundamental questions about how AI systems reshape development practices: To what extent do interactions with AI resemble collaboration with human peers? Where do they diverge, and how do they challenge established assumptions about learning, coordination, and the division of labor in software development?
In this seminar, students will engage in controlled, hands-on programming sessions under two conditions: pair programming with a human partner and pair programming with an AI assistant. Building on these experiences, students will work in small groups of two to three participants to develop research questions and analyze the data collected during the sessions. Each group will be supported by a dedicated advisor and receive regular feedback throughout the semester. The seminar concludes with a short written report and research presentations in which each group presents their results and reflects on their implications.
Kick-Off Meeting: Thursday, 16 April 2026
The seminar takes place Thursdays from 12:00 - 14:00 (~11 sessions in sum)
Participation in all sessions is mandatory.
Requirements: This seminar is open to motivated Bachelor and Master students who are eager to try out AI coding assistants in practice while critically and empirically examining their promised benefits. Previous experience with AI coding assistants is not necessary, but basic software engineering and programming knowledge is.
Places: 13
Aktuelle Themen der Rechtsinformatik by Prof. Dr. Christoph Sorge
Das Seminar "Aktuelle Themen der Rechtsinformatik" ist ein interdisziplinäres Seminar für Informatiker und Juristen.
Ein Interesse an interdisziplinärer Zusammenarbeit ist erwünscht; einige der angebotenen Themen eignen sich besonders dafür, in interdisziplinären Teams bearbeitet zu werden. Für die Anrechnung als Informatikseminar sollte der Schwerpunkt in der Informatik liegen. Es sind auch rein technische Themen möglich, die lediglich Bezug zu juristischen Anwendungen haben.
Die interdisziplinäre Herangehensweise des Seminars ermöglicht es, umfassende und innovative Lösungen für Fragen der Rechtsinformatik zu entwickeln. Das Seminar verbindet Themen wie Privacy Engineering, eJustice und Legal Tech. Privacy Engineering erfordert beispielsweise die Entwicklung und Implementierung komplexer technischer Lösungen wie Anonymisierung, Pseudonymisierung und Privacy-by-Design. Diese technischen Herausforderungen machen das Feld besonders spannend für diejenigen, die gerne innovative und kreative Lösungen entwickeln.
Das Seminar will die Möglichkeit bieten in den interdisziplinären Austausch zu treten und sich mit den technischen und rechtlichen Herausforderungen der Rechtsinformatik auseinanderzusetzen.
Weitere Informationen zum Seminar erfolgen unter: https://www.uni-saarland.de/lehrstuhl/sorge/lehre/sommer-2026.html
Requirements: Es wird erwartet, dass die Teilnehmer in der Lage sind, Vorträgen in deutscher Sprache zu folgen und Ausarbeitungen in deutscher Sprache im Rahmen eines Peer Reviews zu lesen (eigene Vorträge und Ausarbeitungen können in deutscher oder englischer Sprache angefertigt werden).
Places: 7
Audio Devices: Synthesis, Control, & Performance by Paul Strohmeier
In this seminar, students will collaboratively explore the design and implementation of digital music technology. The seminar will build a strong technical foundation in embedded systems and digital audio and provide students with challenges in system design, performance, and aesthetic practice. The final outcome will be a public performance or public art installation.
Students will be evaluated individually on technical skills, and in groups on the successful realization of an artistic vision in a live public setting.
On the technical side, the course covers fundamentals of audio synthesis, embedded audio processing, communication in embedded systems (with a focus on audio-related protocols such as MIDI), sampling of analog and digital sensors, and basic sensor design.
On the design side, students will be challenged to situate these technologies within broader questions of system design, performance design, and artistic practice.
The course is aimed at Computer Science and students with interests in HCI, fine arts, or music. It is also suitable for musicians seeking deeper technical expertise, and for students with a strong technical background (e.g., embedded systems) who wish to apply their skills in a human-centered, creative domain.
Students will leave the course with a clearer understanding of the intersection of art and technology, a solid basis for future work in HCI and related fields, and practical skills relevant to the music technology industry. To support this, the course will include introductions to relevant companies and industry leaders will be invited to the final public presentation of the works created in this course.
The course is relatively compact and requires sustained focus and time commitment from students. It is structured in two phases.
Phase 1 runs from 20.04 to 22.05. During this phase, there will be two meetings per week, which include lectures and presentation of assignments. This phase serves as preparation for Phase 2 and concludes with an exam.
Phase 2 runs from 26.05 to 17.07. During this phase, students will work in groups to develop an instrument or installation. Weekly meetings will provide feedback on progress. Workspace and tools will be provided. Phase 2 concludes with a public presentation of the work and is fully student-led.
Students should expect a workload of at least 15 hours per week. After July 17th, the course is complete. There is no exam after Phase 2.
*
Detailed schedule and additional information can be found here: https://sensint.mpi-inf.mpg.de/instruments.html
Requirements: Students should be proficient in at least one related area:
- Design
- Music or other Performing Art
- Embedded Systems
It is not expected that students have experience in all three areas.
Preferably students have taken either Introduction to HCI or Interactive Systems
Students are encouraged to provide links to design portfolios, music or video hosting services, or git repositories to demonstrate their previous work and expertise
The course is open to beginners with great curiosity, in this case, please provide a strong motivation on why this course is relevant to you.
Places: 16
Augmented Reality for Sports by Dr. André Zenner, Dr. Felix Kosmalla, Muhammad Moiz Sakha, Dr. Florian Daiber, Prof. Dr. Antonio Krüger
In this practical seminar, small groups of students (4) will develop an Augmented Reality (AR) sports application. These applications will either target athletes or spectators as users and aim to solve a certain problem. Both, targeted sport, stake holder, and problem to be solved will be defined in the first weeks of the seminar within the groups with the assistance of the instructors.
Please see the chair's website for more information. https://umtl.cs.uni-saarland.de/teaching/summer-2026/ar-sports.html
Requirements: Since this is a practical seminar, every student should meet the following minimum requirements:
* willingness and ability to create a Developer account for the Meta Quest
* strong programming background
* experience with Unity, C#, and/or AR technology is a plus (but not strictly required)
Places: 12
Computer Algebra by Simon Brandhorst
In this seminar we will look at several topics from the Algebra lecture from an algorithmic and practical point of view. We will address fundamental questions of computer algebra, such as:
How can one efficiently test whether a (large) number is prime?
How can one factor a polynomial with rational coefficients?
How does one compute algorithmically with finite groups?
How to solve linear systems of equations over the integers?
How can one decide whether a nonlinear system of polynomial equations has a (complex) solution?
To answer these and other questions, we will study classical mathematics and modern algorithms and implement them in the programming language Julia.
Requirements: This seminar is aimed at students of mathematics and computer science with an interest in algebra, algorithms, and their applications.
You should already be comfortable with the mathematics underlying the presentation topic — for example, if the topic concerns groups, you should know what a permutation is and when a subgroup is normal.
Places: 14
Computer Architecture by Abdullah Giray Yaglikci and Michael Schwarz
This seminar covers cutting-edge and seminal research papers, focusing on fundamental research problems in the field of computer architecture. Relevant topics include: security and reliability of microarchitecture, memory, and storage, new and emerging paradigms in computer architecture (e.g., data-centric processing), energy efficiency, hardware/software co-design, and fault tolerance.
Requirements: A strong foundation of computer architecture is needed for this seminar.
Places: 8
Data anonymization and how to break it by Ana-Maria Cretu
Anonymization is the main legal paradigm for sharing data while limiting privacy harms. Yet, robust anonymization of individual-level data is very difficult to achieve in practice.
In this seminar, you will first learn what is not anonymous data, through (in)famous examples of anonymization failures. Then, we will turn our attention to modern data sharing systems, including query-based systems, synthetic data, and differential privacy. Finally, we will cover automated approaches for auditing the privacy of these systems, which holds many interesting challenges.
Requirements: Strong interest in data privacy.
Basic knowledge of probabilities and statistics.
(Optional) Having taken a course on machine learning and/or optimization.
Places: 14
Data-driven Understanding of the Disinformation Epidemic (DUDE) by Yang Zhang
Arguably, one of the greatest inventions of humanity is the Web. Despite the fact that it revolutionized our lives, the Web has also introduced or amplified a set of several social issues, like the spread of disinformation and hateful content to a large number of people.
In this seminar, we will look into research that focuses on extracting insights from large corpus of data with the goal to understand emerging socio-technical issues on the Web such as the dissemination of disinformation and hateful content. We will read, present, and discuss papers that follow a data-driven approach to analyze large-scale datasets across several axes to study the multifaceted aspects of emerging issues like disinformation.
During this seminar, the participants will have the opportunity to learn about state-of-the-art techniques and tools that are used for large-scale processing, including, but not limited to, statistical techniques, machine learning, image analysis, and natural language processing techniques.
Requirements: There are no formal prerequisites for this seminar. Despite this fact, it will be helpful if the participants have a basic understanding of machine learning and data mining.
Places: 12
Disagreement in NLP by Frances Yung, Anwesha Das, Vera Demberg
Traditional NLP approaches resolve label disagreements into a single “gold standard,” since disagreements are treated as noise in the data, resulting from the lack of attention or mistakes of the annotators, subjective bias, or insufficient annotation guidelines.
However, recent research highlights that a single gold label may not capture the ambiguity and diversity in language. For subjective tasks such as abuse detection and quality estimation, there is an even greater need for multi-perspective modeling in order to include different viewpoints and improve the robustness and fairness of NLP models.
This seminar explores disagreement in linguistic annotation and perspectivist approaches in NLP, focusing on learning from non-aggregated datasets and multi-perspective evaluation. We will explore the causes of annotation disagreements and strategies to address them. We will discuss current research on modeling diverse viewpoints and the broader implications for AI fairness and inclusion.
We will also discuss how these disagreement-aware models can interact with users in practice, including connections to human-centered AI, adaptive interfaces, cultural difference in visual language processing and cognitive models of interpretation and attention.
Places: 8
Embedded Systems Security by Ali Abbasi
From critical infrastructure to consumer electronics, embedded systems are all around us and underpin the technological fabric of everyday life. As a result, the security of embedded systems is crucial to us.
This seminar, provided by CISPA's EMSEC Lab (https://group.cispa.io/abbasi/), covers research papers addressing various topics in embedded systems security. This might include topics such as instruction profiling (template attack), fault injection and side-channel attacks, firmware static and dynamic analysis, intrusion detection in embedded systems, automotive/space systems security, and fuzzing embedded systems.
More info: https://cms.cispa.saarland/emsecsem_26/
Requirements: Cybersecurity
Places: 8
Fixed Point Theory by Benjamin Kaminski, Tobias Gürtler, Lucas Kehrer, Lena Verscht, Anran Wang
Are you passionate about logic, semantics, and mathematical foundations of computer science? Are you fascinated by self-reference, recursion, and iteration? Then come and study fixed point theory, a central and unifying theme across logic, verification, and theoretical computer science, but also other areas like social sciences and economy.
Fixed points arise whenever definitions, programs, or systems are given recursively. Fixed point theory provides the mathematical tools to reason about existence, uniqueness, approximation, and computation of such solutions, and underpins large parts of semantics and verification.
Topics which we will cover include:
- Fixed points in order theory and lattice theory
- Metric fixed point theory
- Iterative methods and convergence to fixed points
- and many more
The seminar website can be found here:
https://quave.cs.uni-saarland.de/teaching/teaching-ss-2026/fixed-point-theory-ss-2026/
Requirements: The most important requirement: You should really really like math and/or logic. This seminar covers very theoretical work.
The following courses are highly recommended.
Highly Recommended: Programmierung 1, Programmierung 2, Grundzüge der Theoretischen Informatik
Recommended: Semantics; Introduction to Computational Logic; Automata, Games and Verification;
Ideal: Verification,
Places: 8
Formal Methods for AI Alignment: How Do We Make Sure AIs Do the Right Thing? by Kevin Baum, Lisa Dargasz, Felix Jahn, Yannic Muskalla, Andre Steingrüber
AI alignment is, broadly speaking, the task of ensuring that AI systems act in accordance with human values and goals (though what exactly this means remains contested). Many prevalent alignment approaches (such as RLHF and Constitutional AI) suffer from limited interpretability and lack formal guarantees. Bridging these approaches with formal frameworks—both computational and normative—seems tempting, as it promises more rigorous and interpretable alignment techniques.
In this seminar, we will explore a rich repertoire of papers: foundational work on the current state of alignment methods; prima facie suitable formal approaches from computer science (including temporal logics and reward machines) and philosophy (such as deontic logics and theories of reasons); and neuro-symbolic architectures that attempt to combine the strengths of learned systems with formal methods. Participants will have the opportunity to discuss these different strands of research with one another and with members of our interdisciplinary research group.
Requirements: - Genuine interest in questions related to AI alignment and AI safety
- Familiarity with some of the following areas: agentic AI, reinforcement learning, formal logic and formal methods, practical reasoning, and practical philosophy
- Ability and willingness to read and critically discuss English-language research papers
- Openness to interdisciplinary perspectives and active, engaged participation in seminar discussions
Places: 15
GameCraft: Spielmechaniken und Spiele-Prototying by Dr. Pascal Lessel, Prof. Dr. Maximilian Altmeyer (HTW), Prof. Dr. Antonio Krüger
Diese Veranstaltung ist eine hochschulübergreifende Veranstaltung zusammen mit der Hochschule für Technik und Wirtschaft des Saarlandes (HTW). Die Veranstaltung findet auf Deutsch statt.
Studierende werden ein existierendes (und selbst ausgesuchtes) Open Source Spiel auf spielerische Schwächen und Verbesserungsmöglichkeiten hin analysieren (sowohl eigenständig als auch auf Basis von Playtests), Erkenntnisse dokumentieren, Verbesserung auf Ebene der Spielmechaniken konzipieren sowie in das Spiel prototypisch einbauen und durch erneute Playtests zeigen, dass diese tatsächlich auch effektiv sind. Im Verlauf des Semesters, sind mehrere Videodokumentationen anzufertigen und ein Konzept zu erstellen.
Termin: Mittwochs, 10:30–11:30, ab 15.04.2026. (Anwesenheitspflicht). Der Veranstaltungsort ist abwechselnd an der HTW oder der UdS (siehe Webseite).
Es werden zusätzlich Inhalte per Videos vermittelt.
Die Abschlusstermine der Veranstaltung liegen in der vorlesungsfreien Zeit (23.09.2026, 10:30–16:00) und (24.09.2026, 10:00–14:30 nur 20 Minuten in diesem Zeitraum). Die Finalisierung der Verbesserungen erfolgt (je nach eigener Zeitplanung) ebenfalls in der vorlesungsfreien Zeit. Bitte nur eine Priorität vergeben, wenn Sie auch an diesen Terminen anwesend sein können.
Ausführliche Informationen zu den Terminen, Inhalten, Aufgaben und der Benotung finden Sie hier: https://umtl.cs.uni-saarland.de/teaching/summer-2026/seminar-gamecraft-spielmechaniken-und-spiele-prototying.html
Requirements: - Sie sollten sich selbst als Konsument von Spielen sehen (egal ob analog oder digital) und ein Gespür für gute Spielemechaniken haben.
- Sie sollten solide Programmierkenntnisse vorweisen können, um das von ihnen gewählte Open Source Spiel erweitern zu können.
- Sie sollten bereit sein, an allen Veranstaltungsterminen teilzunehmen, unabhängig ob diese an der HTW oder der UdS stattfinden (inkl. der Abschlusstermine in den Semesterferien).
- Sie besitzen einen Laptop, den Sie in den Veranstaltungspräsenzterminen mitbringen können und dessen Hardware es erlaubt, dass Playtests mit Ihrem ausgewählten Spiels durchgeführt werden können.
Places: 8
Generative AI for Creativity by Adish Singla
The course will provide an overview of state-of-the-art research on topics covering generative AI and creativity. The course consists of three main components as follows:
(1) Research papers: During the first half of the semester, students will be assigned research papers for reading and writing reports.
(2) Project: During the second half of the semester, students will work on a project.
(3) Final presentation: At the end of the semester, students will give a presentation on one of the papers and the project.
Requirements: There are no formal requirements; however, please note the following points:
(1) The course is suitable for students who are interested in related topics such as human-computer interaction, educational technologies, psychology, natural language processing, and generative AI.
(2) We are requesting you to provide a short motivation letter explaining the reasons why you are interested in taking this seminar. In this motivation letter, you can also mention any relevant project(s) you have done and any relevant courses you have taken. Please provide this information in the Motivation text box.
Places: 10
Generative AI for Interactive Systems by Ashwin Ram, Ulrike Kulzer, Jürgen Steimle
Recent advances in Generative AI are reshaping how we interact with technology, opening the door to innovative interfaces and new modes of human-computer interaction. This seminar will explore how Generative AI-driven techniques reshape future interfaces. Through a combination of paper readings, discussions, and presentations, students will develop a deeper understanding of key Generative AI techniques within the context of HCI while critically evaluating their role in shaping next-generation user experiences.
Requirements: This is an HCI-centric course. Participants must have completed at least one of the following courses (or an equivalent course at another university): the core lecture “Human-Computer Interaction” or the lecture “Interactive Systems”.
Places: 10
Generative Models in Computer Vision (GMCV) by Jan Eric Lenssen
In this seminar we will discuss the diverse set of paradigms for generative modeling in the area of computer vision. We will cover both seminal works, such as diffusion models, flow matching, and sequential generation paradigms, as well as recent advances in generative modeling for solving inverse problems in 2D and 3D computer vision, such as conditional object and scene generation/reconstruction, novel view synthesis, in-painting, and image super-resolution.
The seminar will consist of an introductory meeting with a lecture at the beginning of the semester introducing the field and distributing papers, and a two-day block course in the semester break covering paper presentations and discussions. Students are expected to read into their assigned paper, the related literature, prepare a talk as well as a paper summary with critical discussion.
Requirements: The student has a solid understanding of of Machine Learning, Computer Vision and feels comfortable with Neural Networks (for example through lectures High Level Computer Vision, Neural Networks: Theory and Implementation, or Machine Learning).
Places: 10
How to build a social computer in the era of LLMs (Interdisciplinary Seminar) by Antonio Krüger, Dimitra Tsovaltzi, Fabrizio Nunnari, Patrick Gebhart, Cornelius König
Computers, mobile phones, and wearables have become an integral part of our daily lives. The user interfaces are designed to interact with factual information, such as health parameters and contextual data. This level of interaction may be adequate for certain applications. However, what if a computer system could understand users like humans do and adapt to their individual social situations to establish a social, human-like interaction? The focus is on social training systems and accesibility /sign-language for different application fields .
This interdisciplinary seminar in Artificial Intelligence and Psychology explores the question of how to make computers social. It gives hands-on experience with concepts and theories that define being social, as well as how humans communicate socially as the basis for building Hybrid AI systems. Therefore, the seminar investigates how these concepts can be transferred into computer models and what is technically feasible and how. As proof of concept, three interactive social computer applications will be created. The project will involve collaboration between students between Computer Science and Psychology, who will design, implement, and evaluate each application.
See previous seminars: https://scaai.dfki.de/teaching/
Requirements: Computer Science students should have knowledge in the areas of AI, HCI, and software design. A bachelor degree in Computer Science (or equal) is required.
Psychology students should have knowledge or high interest in the areas of models of emotions, models of social interaction and requirements.
The main language is English, but communication in German is possible.
Project slides and short progress reports are in English or German.
Places: 16
Interpreting and Analyzing Neural Language Models by Xinting Huang, Michael Hahn
Despite their huge success, neural networks are still widely considered “black-boxes”. In this seminar, we will look into interpretability methods that aim to demystify these models. We will focus on post-hoc interpretability for transformer-based language models, and work on relatively young and burgeoning fields such as Mechanistic Interpretability, which focuses on reverse-engineering model components into human-understandable algorithms. We will read recent papers that involve a diverse set of techniques for interpreting the inner-workings of language models
Requirements: Required: Background in machine learning.
Recommended: Background in natural language processing.
Places: 12
LLMs as Models of Human Sentence Processing by Jiaxin Li, Vera Demberg
Modern large language models (LLMs) have achieved unprecedented success in predicting and generating human language – without being explicitly designed as cognitive models. This raises a fundamental question: to what extent can LLMs serve as computational models of human sentence processing? This seminar dives into this rapidly emerging field at the intersection of NLP, language, and cognition.
We will read papers that compare processing in LLMs against human language processing signatures like eye-movements during reading and data from neurophysiological methods. Furthermore, we will compare types of mistakes that humans vs. LLMs make and read literature on language learning in LLMs vs. human children.
At the end of the seminar, participants will do a final project in which they conduct a project in the topic area of the seminar and write a report about it.
Requirements: Familiarity with language models and transformers; interest in human cognition
Places: 12
Machine Learning for Language Processing by Dietrich Klakow
Deep learning is the predominant machine learning paradigm in language processing. In this seminar we will focus on generic algorithms for language models as well as their application to a variety of languages and other string data eg from bioinformatics.
For more information and the specific list of topics see
Requirements: Strong background in ML
Places: 8
Multimodal Language Understanding by Varsha Suresh, Vera Demberg
Multimodal Language Understanding aims to use information from different sources such as text, speech, images, and gestures, to enhance language processing tasks. As we naturally use multiple forms of communication in our daily interactions, enabling machines to do the same enhances their understanding of human communication. For example, sentiment analysis can be improved by incorporating tone of voice or facial expressions alongside text. In this class, we will explore techniques for modeling multiple modalities, identify tasks that benefit from multi modal input, and discuss the challenges when handling multiple modalities.
For your reference, this is the previous year's course website: https://www.uni-saarland.de/lehrstuhl/demberg/teaching/ws24/25-multimodal-language-understanding.html (this will be updated)
Requirements: This course will include reading, writing, and discussion and is intended for students from Computer Science, Linguistics, and related areas. Knowledge in AI is required, including having taken introductory courses in AI, ML or NLP.
Places: 12
Neural Networks in Brains and Computers by Michael Hahn
We will look into the intersection of machine learning and neuroscience. The main target audience is students with a background in machine learning who are interested in learning about the brain and how it processes information, how artificial neural networks relate to real neurons, how machine learning can help understand the brain better, and how machine learning models may end up aligning with representations found in the brain. The focus will mostly be on vision and language. Thus, this seminar may be particularly interesting if you have a background either in NLP/Computational Linguistics or in Computer Vision. Background in one of the two is certainly enough.
More information: https://lacoco-lab.github.io/courses/brain-2026/
Places: 12
Opportunities and Risks of Large Language Models and Foundation Models by Prof. Dr. Mario Fritz
The advent of Large Language Models (e.g. ChatGPT, Github CoPilot) and other foundation models (e.g. stable diffusion,CLIP) has and will continue to change the way AI/ML applications are developed and deployed. E.g. the behavior and functionality of Large Language Models can be changed entirely by prompting the model, which can be understood as re-programming in plain English.
On the one hand, these models show unprecedented performance and can often be adapted to new tasks with little effort. In particular, large language models like ChatGPT have the potential to change the way we implement and deploy functionality.
On the other hand, these models raise several questions related to safety, security, and general aspects of trustworthiness, that urgently need to be addressed to comply with our high expectations for future AI systems.
Therefore, this seminar will investigate aspects of trustworthiness, security, safety, privacy, robustness, and intellectual property.
This is a lecture in the context of the ELSA - European Lighthouse on Secure and Safe AI: https://elsa-ai.eu
Requirements: Solid understanding of machine learning and deep learning.
Places: 20
Pragmatic processing in Large Language Models by Alexandra Mayn, Vera Demberg
Pragmatics – the study of how language is used in context and how nonliteral meaning is conveyed – is central to human communication. Phenomena such as irony, metaphor, sarcasm, and indirect requests all fall under the umbrella of pragmatics. Given that people use language in nonliteral ways all the time, pragmatic competence is essential for large language models (LLMs) to achieve nuanced language understanding and natural interaction. While LLMs have demonstrated impressive performance across a wide range of tasks, their performance on different pragmatic phenomena, such as recognizing irony and interpreting indirect requests, is more mixed.
In this seminar, we will examine recent research on the pragmatic competence of large language models across different linguistic phenomena. We will also explore whether pragmatic abilities require specific model architectures or fine-tuning strategies, or whether they can be acquired end-to-end given appropriate data and training regimes. In addition, the seminar will survey existing benchmarks and evaluation methodologies used to assess pragmatic abilities of LLMs. At the end of the seminar, participants will do a final project in which they evaluate the performance of LLMs on a pragmatic phenomenon of their choice.
Requirements: Familiarity with LLMs / transformers
Places: 20
Program Synthesis and Compilers by Sebastian Hack
Program synthesis aims to automatically construct correct programs from specifications. Program synthesis has a large potential to automate various parts of building correct-by-design compilers by automatically generating sound code transformations, program analyses, and instruction selector rules.
This seminar focuses on basic program synthesis techniques and looks into applications of synthesis in compilers.
Requirements: For some papers, background in automated reasoning would be good but not a strict requirement. Having passed the compiler construction core course is strongly recommended.
Places: 8
Relational Intelligence: Geometric Deep Learning and the Rise of Agentic Graphs by Rebekka Burkholz
The future of machine learning is relational, where intelligent systems thrive through collaboration, defined by the relationships of agents.
In this seminar, we will first study Graph Neural Networks (GNNs) as a special case of agentic networks to gain insights into design principles for effective agent collaboration and expose novel risks that are introduced by the interconnectedness. We’ll discuss the role of agentic diversity, the risks of oversmoothing, and emerging challenges like adversarial attacks and error propagation cascades.
In the spirit of the seminar, together, let’s gain insights into how we can build robust, collaborative networks and safeguard them from harm as they evolve.
Requirements: The course has no formal requirements but preference will be given to Master students in Computer Science and related fields with prior knowledge in deep learning and a convincing motivation.
More information can be found here:
https://cms.cispa.saarland/rml26/
Places: 14
Research Problems in Machine Learning and Security by Thorsten Eisenhofer and Sven Bugiel
This seminar introduces current research challenges at the intersection of machine learning and computer security. Students will work in small teams on hands-on problems, gaining experience with research methods and learning how to identify and apply relevant ideas from the literature. Over the course of the semester, students will tackle six challenges covering attacks and defenses against machine learning systems, as well as the use of learning-based techniques for practical computer security tasks.
More information: https://cms.cispa.saarland/pine26
Requirements: A strong background in machine learning and computer security is recommended. The course involves independent study, and students should be prepared to fill gaps in their prior knowledge as needed. Since much of the work is carried out remotely on a shared server, students should also have prior experience with, or be willing to independently learn, the necessary tools and workflows for remote development.
Places: 12
Semantic-Driven Music Score Generation by Martin Hennecke, Prof. Dr. Antonio Krüger
This seminar explores how semantic structures extracted from text can control the generation of symbolic musical scores.
In phase 1 (April 16th-May 18th) Students build a minimal semantic-driven music score generation system using LLM-based text representations and Markov chains (e.g. mapping machine-learning–based semantic models or LLM-based transformer embedded vectors to pitch) using Python and LilyPond. Semantically related concepts (“sword” / “dagger”) should result in related musical material, while distant concepts (“sword” / “airport”) should sound correspondingly different.
These systems will be part of a public live lecture concert at Musikfestspiele Saar. Important: As the first system will be part of a public concert, it is important to hold the deadline in Phase 1.
In the second phase (May 28th-June 25th) students each analyse and compare one existing AI–music system or research paper, using their own implementations and The (Un)Answered Question as benchmarks.
In the final phase 3 (June 25th-August 27th) students redesign their approach and develop a more advanced semantic-driven system exploring options of more advanced intelligence and focussing on complex musical structure, polyphony, and motivic development. At the end of semester submission of code and a video documentation (10-15min) about the advanced system is due as the final output.
The seminar combines hands-on programming, experimental composition for a live performance, and research-oriented reflection. It is aimed at students interested in AI, creative systems, and symbolic representation of music.
Regular meeting time: Thursdays, 14:15–15:45
For more information and contact data, see: https://umtl.cs.uni-saarland.de/teaching/summer-2026/semantic-driven-music-score-generation.html
Requirements: - No prior music theory knowledge required, ability to read a music score is helpful.
- Basic programming experience (Python)
- Basic knowledge of or willingness to gain knowledge in AI, machine learning, or creative systems
- Willingness to attend the Lecture Concert (VHS Building near Schloss Saarbrücken) - attending this is mandatory for getting a certificate
- Willingness to work with LilyPond (https://lilypond.org/)
- Willingness to work collaboratively and experiment
- Availability in lecture free period until end of August (finalization of system, video documentation)
- Reliability and dependability
Places: 8
Special Topics in Deep Reinforcement Learning by Verena Wolf, Timo P. Gros
Reinforcement learning (RL) is a popular technique to solve decision-making problems. In combination with artificial neural networks, it has achieved huge successes in challenging domains such as mastering Atari Games, Chess, or GO.
In this seminar, the participants will learn the theory and practice of the most important concepts of deep reinforcement learning (DRL) and details of popular DRL algorithms. Moreover, it will cover special topics such as exploration techniques in DRL, neuro-symbolic RL, or Generalization in RL.
The participants will give presentations and also implement some of these algorithms in small teams to solve the ProgGrid Traffic Gym (PGTG) benchmark. The practical project has been highly appreciated in earlier editions of the seminar.
Requirements: Only participants who successfully participated in the lecture 'Reinforcement Learning' may apply.
Places: 12
Sports Analytics by Jun.-Prof. Dr. Pascal Bauer
This course introduces sports analytics through interdisciplinary collaboration between students in project-based teams. Students apply data science and machine learning methods to real-world sports data, including complex spatio-temporal tracking data. The course is organized into three blocks: sports
analytics lectures (theory and literature), applied projects focusing on student-driven problem-solving concepts, and results presentation. Through hands-on analyses, collaborative work, and presentations, students integrate theoretical knowledge with practical skills for research and applied sports analytics.
When? Fridays 10:15-11:45
More detailed information about the course structure can be found here:
Requirements: Basic programming skills (Python or R), fundamental knowledge of statistics, and prior experience with a data science or machine learning project.
Places: 20
Topics in Inferential Statistics (with R) by Emilia Ellsiepen, Vera Demberg
This seminar examines approaches to statistical inference that extend beyond the classical linear model. Emphasis is placed on understanding how inferential conclusions are affected by model complexity, data imperfections, and modern estimation techniques, alongside practical implementation in R.
Students are expected to be comfortable with linear and generalized linear models, including mixed-effects models. Possible topics include inference with missing data via multiple imputation, generalized additive models, mixture models and latent structure, multiple testing and false discovery rate control, penalized regression methods (e.g. ridge and lasso), and tree-based methods such as random forests.
Throughout the seminar, methods are evaluated not only for their predictive performance but also for the validity, interpretation, and robustness of the inferences they support.
Students will engage with the statistical literature and present a topic or complete an applied project using real data and R.
Requirements: Only students who successfully participated in Statistics with R may apply
Places: 10
Topics in Privacy-Preserving Machine Learning and Optimization by Andrew Lowy, Sebastian Stich
This seminar explores modern research in differentially private (DP) optimization and machine learning, a framework that enables learning from sensitive data while providing rigorous privacy guarantees. We will study core concepts in differential privacy and optimization, and examine recent advances in DP optimization, with emphasis on fundamental privacy-accuracy tradeoffs.
In addition, the seminar will cover recent work on machine unlearning, which addresses the problem of efficiently removing individuals’ data from trained models. We will discuss how unlearning relates to, but is distinct from, differential privacy, and study efficient algorithms and accuracy guarantees for unlearning.
Students will read and present current research papers, develop skills in understanding theoretical machine learning literature, and identify open research directions at the intersection of optimization, privacy, and learning theory.
More info: https://cms.cispa.saarland/ppmlo_26/
Requirements: Students should have a solid command of probability and multivariable calculus. Comfort with mathematical proofs is strongly recommended.
Places: 14
Trusted AI Planning (TAIP) by Daniel Höller
AI planning agents need to take action decisions towards a long-term objective. Learned action policies, in particular neural networks, that map environment states to actions are gaining ever more popular for this purpose. Yet such learned policies come without any inherent guarantees regarding desirable properties such as safety. The seminar covers recent research on quality assurance methods, including verification, testing, and re-training of learned action policies.
Requirements: Students must have passed the AI core course or the Trusted AI Planning specialized course.
Places: 16
Twelve test generators for twelve programming languages by Alexi Turcotte + Andreas Zeller
In this seminar, we will study and produce _test input generators_ for twelve different statically typed programming languages, including Java, Go, Rust, Kotlin, Scala, Haskell, ML, and other languages of your choice.
At CISPA, we have developed test generators that produce test inputs in highly complex formats and languages. Recently, we have been successful in getting these generators to produce lots of valid and diverse C programs. We can feed these into a C compiler and see what happens. (Some compilers actually crash.)
Our method requires (1) a _grammar_ of the programming language in question (these are always already available), and (2) a set of _constraints_ that define rules such as definition before use, proper scoping, or basic type correctness. We already have such constraints in place for the C language. Since the rules do not vary much across languages, adapting the C rules to another programming language may not require more than a few days.
In this seminar, we will study different programming language concepts every week, with each participant focusing on applying that concept to the generator for their chosen language, and then together discussing how these concepts are realized. At the end of the seminar, your generator may become part of a high-profile scientific publication we plan to write in Summer – with you as a co-author!
Requirements: To be successful in this seminar, you need:
* basic knowledge of grammars, as you would typically have gained in introductions to computer science
* a good understanding of the semantics of programming languages, such as identifier usage, scoping, typing, and more.
Constraints are specified in Rust, so knowledge of Rust will be helpful. However, you will be able to learn Rust during the seminar.
Successful participation in "Security Testing" or equivalent knowledge of structured fuzzing is a plus.
Places: 12
Van Horn to Fuchsia: Capability-based Access Control by Sven Bugiel
Capabilities as an access-control paradigm have been around for nearly 6 decades. Still, until recently, they were a little-known access-control primitive and were not deployed in end-user devices. However, with the advent of CHERI and Google Fuchsia, they recently re-emerged. In this seminar, we will discuss how capability systems evolved over the decades and which properties could benefit modern systems. Specifically, mobile, appified systems like Android and iOS.
Since capabilities are a long-established area of research, the selected papers include current and historical works.
CISPA CMS: https://cms.cispa.saarland/capac26/
Requirements: Basics of access control and operating system security are required (e.g., passed the Foundations of Cybersecurity Lectures or the Core Lecture Security).
Places: 8