Tufa Talks episode 2 - Mattia Rigotti (IBM Research Zürich)

🌟 Join us for the second episode of our engaging Tufa Talks series held at our office! We're a small, independent research group dedicated to fundamental AI research, aiming to share groundbreaking machine learning insights while fostering open discussions and collaborations among ML researchers in Zürich.

🎤 Featured Speaker: Mattia Rigotti, a Senior Research Scientist at IBM Research Zürich and a PhD alumnus from Columbia University in Neuroscience. Mattia's research elegantly bridges machine learning and neuroscience, covering areas such as LLM reasoning, visual foundation models, and the intricate geometry of neural representations in the brain.

🔍 Known for his innovative work on mixed selectivity in the prefrontal cortex, he reveals how high-dimensional neural representations can promote flexible cognition. His latest research investigates whether reasoning in LLMs can occur without relying on reinforcement learning. By leveraging insights from cognitive psychology, Mattia explores how equipping a base model with a set of "cognitive tools" — modular reasoning operations executed by the LLM itself — can effectively unlock structured, multi-step reasoning.

Agenda:

• 📍 Location: Tufa Labs, Nüschelerstrasse 30
• 🍕 Post-Talk: Enjoy pizza, drinks, and networking opportunities!

💡 Note: Registration is required, and spots are limited, so ensure you secure your place! If selected, you will be notified by the Friday prior to the event. We can't wait to see you there!