Since the early days of LLMs, we have been working to improve the reasoning capabilities of the models.
Our focus includes developing analytical tools to measure and evaluate these capabilities, as well as new methodologies to augment reasoning.
In particular, we are actively working on dynamic reasoning, tool-calling and agentic systems.
We specialize in the fundamentals of AI, focusing on deep learning analysis, neural network learning complexity, and hierarchical/curriculum learning. Our research includes hardness and separation results comparing basic, overparameterized, and chain-of-thought models. Additionally, we have extensive experience in high-dimensional statistics applied to unsupervised learning, specifically dimensionality reduction and community detection
Our work here focuses on discovery research. While cardiovascular diseases remain the leading cause of death worldwide, progress in diagnosis and treatment has largely plateaued. AI is a natural candidate to drive innovation in the field; however, advanced methods based on scientific AI and reasoning are required to navigate complex, low-volume data regimes. We are addressing these challenges through our SwissCardIA collaboration
For decades, this field has defined the measures and fundamental limits for communication and learning systems. We focus on developing algorithms that reach these limits and on establishing new metrics tailored to modern machine learning environments
