Mirco Mutti

I'm a postdoctoral researcher at the Technion - Israel Intitute of Technology, where I work with Aviv Tamar in the Robot Learning Lab. Previously, I obtained my PhD in Data Science and Computation at Politecnico di Milano advised by Marcello Restelli.

My main interest is in reinforcement learning. The aim of my research is to build the fundamentals of reinforcement learning generalization for training artificial agents that can solve a variety of complex sequential decision making problems in the real world. On this line, I contributed advances on unsupervised pretraining, meta RL, partial observability, general utilities, RLHF, imitation learning, and inverse RL among others. My research has been awarded a best paper award at ICML 2022.

Feel free to drop an email for inquiries and questions. You can find my (mostly) up to date CV below. Turn to my Scholar or social profiles to catch up with more recent agenda.

Email  /  CV  /  Google Scholar  /  LinkedIn  /  Bluesky  /  Twitter

• Traveling to San Diego for NeurIPS 2025. We are going to present three posters (1, 2, 3). Our paper on State entropy regularization for robust RL has been awarded an oral presentation.
• I had the pleasure to talk about my current research on RL generalization at the Princeton Reinforcement Learning Lab.
• Our paper Blindfolded experts generalize better has been awarded a best paper at the EXAIT workshop at ICML 2025.
• Heading to Vancouver for ICML 2025. We are going to present A classification view of meta learning bandits and Enhancing diversity in parallel agents plus a pair of workshop treats.
• I recently toured northern Italy - MaLGa and IIT in Genova, Bocconi, University of Verona - to talk about my recent works on reinforcement learning: From theory to practice.
• Our paper A theoretical framework for partially-observed reward states in RLHF has been accepted at ICLR 2025.
• Check out our new preprint Reward compatibility: A framework for inverse RL to read about our journey into the theoretical foundations of inverse reinforcement learning.
• Our paper How does inverse RL scale to large state spaces? A provably efficient approach has been accepted at NeurIPS 2024.
• Our paper The limits of pure exploration in POMDPs: When the observation entropy is enough has been accepted at brand-new RLC conference.
• Four papers accepted at ICML 2024. They cover meta RL, inverse RL, geometric active exploration, and pure exploration in POMDPs. See you in Vienna!
• I gave a talk at the VANDAL lab on Unsupervised reinforcement learning in Turin.
• My PhD thesis received an honorable mention for the best PhD thesis on artificial intelligence by the AIxIA
• I gave a talk at the ETH LAS group and AI Center on (Non)convex reinforcement learning in Zurich.
• Happy to announce I am joining the Technion as a postdoctoral researcher from September 2023. I will work on reinforcement learning from theory to practice with Aviv Tamar.
• I succesfully defended my PhD thesis on March 2023.
• Our paper A tale of sampling and estimation in discounted reinforcement learning has been accepted at AISTATS 2023 with an oral presentation.
• I have been invited to give a talk as a Rising stars in AI at KAUST.
• I have served as co-program chair for EWRL 2022 we organized in Milan.
• Grateful that our paper on The importance of non-Markovianity in maximum state entropy exploration has been recognised with the outstanding paper award at ICML 2022.

Template here.