Large Language Models: Foundations and Applications (PhD level)
Period: 2025
Contact: Matteo Magnani
Teacher: Lucio La Cava, University of Calabria
Description of the course
Large Language Models (LLMs) are transforming research and innovation, reshaping how we generate, interpret, and utilize knowledge. This 10-hour course provides a comprehensive overview of Large Language Models, from foundational theory to applications in trending research directions.
The first part of this course introduces the core concepts behind LLMs, focusing on the underlying architectures (e.g., recurrent networks, attention mechanisms, and Transformers), training and inference techniques (e.g., pre-training, fine-tuning, preference alignment, and prompting), and their promising language understanding and generation capabilities. This part also includes a discussion on the societal and ethical impact of this technological shift.
The second part explores challenging and emerging topics currently investigated within the Artificial Intelligence and Data Science Laboratory of the University of Calabria, including methodologies and resources for detecting and attributing AI-generated content, using LLMs for challenging tasks such as lexical entailment, the possibility of mimicking human personality traits via LLMs, and morality aspects in AI-generated or manipulated content.
Throughout this course, the audience is expected to gain both theoretical understanding and exposure to promising research directions, developing a crucial understanding of LLMs’ capabilities and limitations, and exploring their potential in real-world applications.
Biography of the teacher
Lucio La Cava is an Assistant Professor (RTD-A) at the Department of Computer Engineering, Modeling, Electronics, and Systems Engineering (DIMES) of the University of Calabria, Italy, funded under the PNRR Project SERICS (PE00000014), EU-NextGenerationEU. Previously, he got a Ph.D. cum laude in Information and Communication Technologies at the DIMES Department of the University of Calabria, with a thesis on “Graph Mining and Multimodal Representation Learning for Decentralized Socio-Economic Domains.” He was a visiting Ph.D. student at the IT University of Copenhagen working in the NEtwoRks, Data, and Society (NERDS) group. He is a member of the Machines, Languages, and Networks Team with the Artificial Intelligence and Data Science Laboratory of DIMES. His research interests include Natural Language Processing (with particular focus on Large Language Models), Multimodal Representation Learning (including Graph Representation Learning and Computer Vision), and Graph Mining.
Time plan
The course is given over two days:
- Wednesday, May 21 (9:00-12:00, 13:00-17:00)
- Thursday, May 22 (9:00-12:00)
Credits and assessment
Students must actively attend the meetings to obtain 0.5 credits. Additional credits can be obtained through follow-up work on the topic, to be agreed with the teacher.
Expected level and prerequisites
The course is for PhD students at all stages of their education, but also open to Master’s students and researchers interested in getting an intensive introduction to the topic. Credits are however only available for PhD students.
Registration
The course is free but the number of available places is limited. Please register by sending an email to matteo.magnani@it.uu.se