Seminar Data-Science I (Methods)
CS 721 Master Seminar (M. Sc. Wirt. Inf., M.Sc. MMDS, Lehramt für Gymnasien)
Lecturer | Abigail Hayes, Jana Jung, Marlene Lutz, Jens Rupprecht |
Course Format | Seminar |
Offering | HWS/ |
Credit Points | 4 ECTS |
Language | English |
Grading | Written report with oral presentations |
Examination date | See schedule below |
Information for Students | The course is limited to 16 participants. Please register centrally via Portal2. |
Contact
For administrative questions, please contact Abigail Hayes.

Abigail Hayes
L 15, 1–6
3. OG – Raum 324
68161 Mannheim
Course Information
Course Description
In this seminar, students perform scientific research, either in the form of a literature review or by conducting a small experiment, or a mixture of both, and prepare a written report about the results. Topics of interest focus around a variety of problems and tasks from the fields of Data-Science, Network Science and Text Mining.
Previous participation in the courses Network Science and Text Analytics are recommended.
Objectives
Expertise: Students will acquire a deep understanding of the research topic. They are expected to describe and summarize a topic in detail in their own words, as well as to judge the contribution of the research papers to ongoing research.
Methodological competence: Students will develop methods and skills to find relevant literature for their topic, to write a well-structured scientific paper and to present their results.
Topics
This seminar will be split into four main topic blocks. Every student will be assigned a research paper from only one of these blocks to work on. Yet, it is expected that students also actively participate in the discussion of papers from other topic blocks after they have been presented.
The four topics we are going to discuss in the HWS 2025 are:
- Fairness in Graph Neural Networks. Fairness is not only important when considering Large Language Models (LLMs) and textual data. As Graph Neural Networks (GNNs) are increasingly used in real-world applications, including for recommendations and social analysis, ensuring fairness becomes essential. Research has demonstrated that GNNs can be biased by data, vulnerable to fairness attacks, or limited by architectural choices. We will understand how fairness is defined in graph settings and the methods being proposed to make GNNs more equitable and robust.
- Parameter Selection for LLM Inference. This topic explores the impact of inference configurations—such as temperature, top-k, and top-p sampling—on the output of LLMs. As the deployment of LLMs becomes widespread in academia, industry, and society, understanding how decoding choices shape model behavior is critical for ensuring output quality. The papers examine foundational and state-of-the-art research, including sampling algorithms and contrastive decoding methods. It discusses challenges such as hyperparameter sensitivity, task dependency, and the limitations of current evaluation metrics.
- A Thorough Examination of Decoding Methods in the Era of LLMs
- DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models
- Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs
- Balancing Diversity and Risk in LLM Sampling: How to Select Your Method and Parameter for Open-Ended Text Generation
- Decision-Making of LLM Agents. Language agents powered by Large Language Models (LLMs) are becoming increasingly important as decision-making tools across a wide range of applications. As their influence grows, so does the need to improve their ability to make accurate, consistent, and context-aware decisions. This project explores various strategies to enhance the decision-making capabilities of LLM agents, focusing on advanced prompting techniques and reinforcement learning approaches. Improving these capabilities is essential to ensure that LLMs can serve as reliable collaborators and support systems in high-stakes and complex decision-making environments.
- Multilinguality.
- How do languages influence each other? Studying cross-lingual data sharing during LM fine-tuning
- Do Llamas Work in English? On the Latent Language of Multilingual Transformers
- Do Multilingual LLMs Think In English?
- Do Large Language Models Have an English Accent? Evaluating and Improving the Naturalness of Multilingual LLMs
Schedule
The schedule below is preliminary, dates are subject to change.
Registration period until 01.09.2025, 23:59 via Portal2 Kick-off meeting 10.09.25, 9:00–9:45
L 15 1-6, room 314/
315 General information Drop-out until tba 1st Presentation Date 13.10. or 20.10.25 , 10:15–13:15
L 15 1-6, room 314/
315 Presentations 2nd Presentation Date 27.10. or 03.11.25, 10:15–13:15
L 15 1-6, room 314/
315 Presentations Submission deadline 28.11.2025, 23:59 Written report Registration
Please register via Portal2.