I am an Assistant Professor at Sapienza University of Rome and a member of the PINlab and ELLIS Society. My research centers on AI agents built with VLMs and LLMs, focusing on their safety, collaboration, and memory across both real and digital worlds. This builds on my background bridging 3D/4D computer vision and robotics to reconstruct and re-simulate the dynamic world. I am an NVIDIA Academic Grant recipient and Principal Investigator of multiple research initiatives, including a Sapienza seed grant to found the startup advAIse, where I’m serving as Co-founder and CTO.
Jun 2026 Received the Best Reviewer Award at CVPR 2026.
Jun 2026 Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models is accepted at ECCV 2026.
Jun 2026 Robots Ask the Way: Communication-Enabled Social Navigation is accepted at IROS 2026.
Feb 2026 Video Unlearning is accepted at ICLR 2026.
Dec 2025 Received the Best Reviewer Award at NeurIPS 2025.
Nov 2025 Xuanhao Fan started his PhD under my supervision.
Oct 2025 Human Motion Unlearning is accepted at AAAI 2026.
Sep 2025 NVIDIA Academic Grant “LLMs Can teach your robot to speak”.
Jul 2025 MonSTeR is accepted at ICCV 2025.