Part of the
4TU.
NIRICT
TU DelftTU EindhovenUniversity of TwenteWageningen University
4TU.
NIRICT
Close

4TU.Federation

+31(0)6 48 27 55 61

secretaris@4tu.nl

Website: 4TU.nl

Program 4TU.NIRICT Community Day, Feb 5, 2025

Tuesday, 4 February 2025

(Room: Eem-Lek and Foyer for lunch and breaks)  

12:30: Walk-in and Lunch  

13:15- 13:30: Opening and Welcome (Suzan Bayhan

13:30-15:00: GenAI for ICT: Developments, opportunities, and threats (moderator: Tarek Alskaif) 

15:00-15:30: Networking and Coffee break 

15:30-16:15: Group-work session (moderators: Marc Geilen, Tarek Alskaif, Suzan Bayhan)  

16:15- 16:30: Coffee break 

16:30-17:00: 4TU.NIRICT Funded Projects (moderator: Vasilios Andrikopoulos

17:00 –18:00: Closing & Drinks 

 More information on the talks:  

 Bridging the Retrieval and Reasoning Gap in Large Language Models 
Avishek Anand, TU Delft  

Abstract: 
The advent of large language models (LLMs) has revolutionized information seeking, ushering in a new paradigm for knowledge retrieval and reasoning. Despite their capabilities, LLMs remain prone to hallucination, generating plausible yet nonfactual content. This raises critical concerns about their reliability in real-world information retrieval (IR) systems, spurring intensive research to detect and mitigate such issues. In this talk, we will explore the circumstances under which LLMs are likely to hallucinate or provide incorrect answers. We will then delve into two critical research gaps that hinder the reliability of LLMs for information-seeking tasks: the retrieval gap and the reasoning gap. I will present recent advancements in meta-prompting techniques, showcasing how the careful selection of representative examples from extensive datasets can significantly enhance LLMs’ reasoning capabilities. Finally, we will discuss a new family of approaches, termed online relevance estimation, which addresses the retrieval gap in retrieval-augmented generation (RAG) systems, offering a pathway to more robust and trustworthy LLM-driven information systems. 

Bio: 
Avishek Anand is an Associate Professor in the Web Information Systems (WIS) group within the Software Technology department at Delft University of Technology (TU Delft). He also serves as the director of the Research, Engineering, and Infrastructure Team (REIT) in the Department of Software Technology. His research focuses on developing intelligent and transparent machine learning techniques to assist users in finding relevant information, with a particular emphasis on Explainable Information Retrieval. He holds a PhD in Computer Science from the Max Planck Institute for Informatics, Saarbrücken. Prior to joining TU Delft, he was an Assistant Professor in Information Retrieval at Leibniz University Hannover and a visiting scholar at Amazon Search. His research has been supported by organizations such as Amazon Research Awards, Schufa GmbH, the German Federal Ministry of Education and Research (BMBF), and the EU Horizon 2020 program. Avishek has delivered numerous keynotes, tutorials, and invited talks at leading information retrieval conferences and summer schools. His research contributions have received multiple best paper awards at top-tier AI conferences. 

 
[Gender] Bias in [Multilingual] Language Technologies 
Eva Vanmassenhove, Tilburg University 

Abstract: 
Multilingual language technologies promise to bridge linguistic divides, enabling global communication and access to information. However, these systems often struggle with biases— among other gender and linguistic biases—that manifest in machine translations, text generation, and speech processing. From fairly gender-neutral languages being forced into gender classifications to the marginalization of lower-resource languages, AI models frequently reflect and reinforce societal and linguistic inequalities. In this talk, we will explore why some of these biases emerge and discuss the potential implications for users. We will furthermore discuss some of the challenges in building fair and inclusive multilingual AI, present real-world examples of biased outputs, and explore some strategies for mitigating these biases. At the same time, we want to reflect on the potential opportunities. How can we create or work towards language technologies that better respect linguistic diversity and gender inclusivity? 
 
Bio: 
Eva Vanmassenhove is a researcher specializing in Machine Translation and Language Technology, with a strong focus on tackling gender and algorithmic biases in translation systems. She earned her PhD from Dublin City University and now serves as an assistant professor in the Department of Cognitive Science and Artificial Intelligence at Tilburg University (TiU). At TiU, she contributes to the Computation and Psycholinguistics Research unit and the Inclusive and Sustainable Machine Translation Research Line. Her work aims to enhance machine translation by addressing biases, especially in gender representation, while preserving linguistic richness.  

 
Cloud Deployment of Machine Learning Applications for Image Guided Therapy   
Danny Ruijters, Philips and TU/e 

Bio: 
Danny Ruijters is a principal scientist at Philips, with a track record of creating prototypes for interventional patient treatment, evaluating those prototypes in a live clinical setting, taking into account the technical, clinical and regulatory aspects. He is also a part-time professor in the area of data driven value-based healthcare in image guided therapy. This comprises the development of intelligent and context aware systems that optimize the data gathering and application in minimally invasive treatment. Particularly, the translation of large population datasets to the individual patient and vice-versa, and the direct application during image guided therapy is part of his research focus.