Archive
/
INF Seminars
/
INF_2025_03_28_RamonRuizDolz
USI - Email
Università
della
Svizzera
italiana
INF
IDSIA Seminar
Browser version
Argumentation: The Missing Piece for Natural Language Reasoning in LLMs
Host: Prof. Fabio Crestani
Venerdì
28.03
USI East Campus, Room D0.02
15:30 - 16:30
Ramon Ruiz-Dolz
University of Dundee, UK
Abstract: Since the release of the latest versions of popular LLMs such as o3, Gemini 2.0 Flash Thinking, or DeepSeekV3-R1, the race for Artificial General Intelligence (AGI) has put its focus on the “reasoning” capabilities of these models. Advances in the “reasoning” capabilities of LLMs are measured and quantified with the use of benchmarks including a wide set of tasks and challenges that address reasoning from different perspectives (e.g., commonsense, mathematical, or multi-hop reasoning). It is interesting, however, to see that none of these tasks involve argumentation, the natural form of reasoning in natural language. In this seminar, Ramon will briefly introduce the nature of the “reasoning” capabilities of LLMs, covering some of the tasks included in the benchmarks. Then, fundamental concepts of argumentative reasoning will be presented, pointing out the main differences between reasoning in natural language with arguments, and the tasks used to benchmark reasoning. Finally, the seminar will end with a discussion on the most important open challenges that will push forward the boundaries of reasoning in LLMs.
Biography: Ramon Ruiz-Dolz is a Lecturer in Computing at the University of Dundee (UK). His main areas of research are Computational Argumentation and Natural Language Processing. His current research focuses on the analysis of the natural language reasoning capabilities of LLMs, and the integration of concepts from argumentation theory into NLP algorithms to improve their explainability and to develop critical thinking skills. Ramon obtained his PhD in Computer Science from the Universitat Politècnica de València (Spain).