About Me
I am a Ph.D. student at Politecnico di Milano, working on Reinforcement Learning and Robotics, under the supervision of Prof. Alberto Maria Metelli. Recently, I also spent time at UC Berkeley working on action chunking with diffusion policies under the supervision of Prof. Sergey Levine.
In 2021, I received a B.Sc. in Computer Science and Engineering from Politecnico di Milano (with honors), and in 2023, I earned a M.Sc. in Computer Science and Engineering from Politecnico di Milano (with honors), graduating one semester early. During my M.Sc., I attended the prestigious ASP honor program, learning the nuances of innovating. After graduation and before starting my Ph.D., I worked as a Machine Learning Engineer at ML3.
In my free time, I enjoy sports, particularly skiing ⛷️ and playing tennis 🎾.
Download my Curriculum Vitae (updated August 2026).
Research Interests
I aim to bring capable AI agents into the physical world. From a robot-learning perspective, I study how robots can acquire robust behaviors from human data in an efficient way, with a particular focus on behavioral cloning policy pre-training. I am especially interested in methods that address the key challenges of the field, such as test-time distribution shift and the limited availability of robotic data compared with the vast datasets used to train LLMs.
Selected Publications
See my Google Scholar for a full list.Why Does Action Chunking Improve Behavioral Cloning Performance in Robotic Control?
Filippo Lazzati, Kyle Stachowicz, William Chen, Alberto Maria Metelli, Andrew Wagenmaker, and Sergey Levine.
Pre-print.
[Link]Imitation Learning as Return Distribution Matching
Filippo Lazzati and Alberto Maria Metelli.
ICLR 2026.
[Link] [arXiv]How does Inverse RL Scale to Large State Spaces? A Provably Efficient Approach
Filippo Lazzati, Mirco Mutti and Alberto Maria Metelli.
NeurIPS 2024.
[Link] [arXiv]Towards Theoretical Understanding of Inverse Reinforcement Learning
Alberto Maria Metelli, Filippo Lazzati and Marcello Restelli.
ICML 2023, Oral.
[Link] [arXiv]