Scholar
Matteo Bettini
Google Scholar ID: hcvR_W0AAAAJ
PhD Candidate, University of Cambridge
Citations & Impact
Citations
406
H-index
8
i10-index
8
Publications
17
Co-authors
12
list available
Publications
Resume (English only)
Academic Achievements
- Published several papers including 'ARE: Scaling Up Agent Environments and Evaluations' (preprint), which introduces Meta Agents Research Environments (ARE) as a research platform for scalable creation of environments, integration of synthetic or real applications, and execution of agentic orchestrations; 'Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning' (ICML conference paper), proposing a method called Diversity Control to control behavioral diversity in MARL; and 'BenchMARL: Benchmarking Multi-Agent Reinforcement Learning' (JMLR journal article), presenting a library for benchmarking MARL using TorchRL.
Research Experience
- During his PhD, he studied heterogeneity in multi-agent and multi-robot systems and developed the VMAS simulator at the Prorok Lab. He also joined PyTorch at Meta, where he created BenchMARL and helped develop TorchRL. For his master's, he investigated transport network design for multi-agent routing.
Education
- PhD in Computer Science from the University of Cambridge in 2025; MPhil in Advanced Computer Science from the same university in 2021; BEng in Computer Engineering from Politecnico di Milano in 2020.
Background
- Research interests include reinforcement learning, multi-robot systems, LLM agents, heterogeneous multi-agent learning and coordination, and graph neural networks. Currently working at Meta on training LLM agents for long-horizon tasks using reinforcement learning.
