Pietro Mazzaglia

Pietro Mazzaglia

Hi, I am Pietro Mazzaglia and I am a Senior AI Researcher at Qualcomm AI Research, Amsterdam.
The long-term goal of my research is to build embodied AI agents that discover and learn how to behave in the environment from interactions data.

Google Scholar GitHub X LinkedIn

About

  • I grew up in Catania, a sunny city next to a volcano in the east coast of Sicily, Italy.
  • During my bachelor, I lived and studied at the superior graduate institute Scuola Superiore di Catania, where I was a scholar while attending the BSc Computer Engineering course at the University of Catania.
  • For my master degree, I moved to Manchester, UK, where I studied Artificial Intelligence at The University of Manchester.
  • I did my PhD at Ghent University, Belgium, supervised by Prof. Bart Dhoedt and Dr. Tim Verbelen. During my PhD, I also interned at Qualcomm AI Research (Amsterdam, Netherlands), Dyson Robot Learning Lab (London, UK) and ServiceNow Research (Montreal, Canada).

Research interests

Embodied AI World Models Multimodal Agents Reinforcement Learning Robotics Active Inference

Selected publications

Here's a selection of my research contributions. For an exhaustive list, see my Google Scholar profile.


Hybrid Training for Vision-Language-Action Models
Hybrid Training for Vision-Language-Action Models
Pietro Mazzaglia, Cansu Sancaktar, Markus Peschl, Daniel Dijkman
Preprint
Paper
Focusing on What Matters: Object-Agent-centric Tokenization for Vision Language Action Models
Focusing on What Matters: Object-Agent-centric Tokenization for Vision Language Action Models
Rokas Bendikas*, Daniel Dijkman*, Markus Peschl, Sanjay Haresh, Pietro Mazzaglia
CoRL 2025
Paper
GenRL: Multimodal-foundation world models for generalization in embodied agents
GenRL: Multimodal-foundation world models for generalization in embodied agents
Pietro Mazzaglia, Tim Verbelen, Bart Dhoedt, Aaron Courville, Sai Rajeswar
NeurIPS 2024 Outstanding Paper @ ICML workshop 2023
Paper Code Project
FOCUS: Object-Centric World Models for Robotics Manipulation
FOCUS: Object-Centric World Models for Robotics Manipulation
Stefano Ferraro*, Pietro Mazzaglia*, Tim Verbelen, Bart Dhoedt
Frontiers in Neurorobotics 2025 Best Paper @ RSS workshop 2023
Paper Code Project
Choreographer: Learning and Adapting Skills in Imagination
Choreographer: Learning and Adapting Skills in Imagination
Pietro Mazzaglia, Tim Verbelen, Bart Dhoedt, Alexandre Lacoste, Sai Rajeswar
ICLR 2023 Spotlight (top 25%)
Paper Code Project
Information-driven Affordance Discovery for Efficient Robotic Manipulation
Information-driven Affordance Discovery for Efficient Robotic Manipulation
Pietro Mazzaglia, Taco Cohen, Daniel Dijkman
ICRA 2024
Paper Project
Redundancy-aware Action Spaces for Robot Learning
Redundancy-aware Action Spaces for Robot Learning
Pietro Mazzaglia*, Nicholas Backshall*, Xiao Ma, Stephen James
RA-L 2024
Paper Code Project
Mastering the Unsupervised Reinforcement Learning Benchmark from Pixels
Mastering the Unsupervised Reinforcement Learning Benchmark from Pixels
Sai Rajeswar*, Pietro Mazzaglia*, Tim Verbelen, Alexandre Piché, Bart Dhoedt, Aaron Courville, Alexandre Lacoste
ICML 2023 Oral
Paper Code Project
Curiosity-Driven Exploration via Latent Bayesian Surprise
Curiosity-Driven Exploration via Latent Bayesian Surprise
Pietro Mazzaglia, Ozan Çatal, Tim Verbelen, Bart Dhoedt
AAAI 2022 Oral
Paper Code Project Blog
Contrastive Active Inference
Contrastive Active Inference
Pietro Mazzaglia, Tim Verbelen, Bart Dhoedt
NeurIPS 2021
Paper Code Project Blog
The Free Energy Principle for Perception and Action: A Deep Learning Perspective
The Free Energy Principle for Perception and Action: A Deep Learning Perspective
Pietro Mazzaglia, Tim Verbelen, Ozan Çatal, Bart Dhoedt
Entropy 2022 Best Paper Award
Paper