Benchmarking World-Model Learning
A. Warrier, D. Nguyen, M. Naim, Moksh Jain, Y. Liang, K. Schroeder, Cambridge Yang, Joshua B. Tenenbaum, Sebastian Josef Vollmer, Kevin Ellis, Z. Tavares
People
Michelangelo Naim
Research Scientist, Basis Research Institute
Michi is a research scientist at Basis, working at the intersection of AI, physics, and math. His work focuses on computational theories of intelligence, including reasoning, learning, and decision-making, alongside open-source software.
About
I am a research scientist at Basis, where I specialize in the intersection of AI, physics, and math. My work focuses on developing computational theories of intelligence while contributing to and maintaining open-source software.
I am deeply committed to AI safety and its responsible advancement. I am particularly interested in mechanistic interpretability and inner alignment in neural networks, and I aim to make AI more understandable, believing this will help us learn and grow faster.
Previously, I was an ICoN Postdoctoral Fellow at MIT, working with Guangyu Robert Yang using AI to model the brain. I completed my Ph.D. in Theoretical Neuroscience at the Weizmann Institute of Science under the supervision of Misha Tsodyks, during which I spent time at the Institute for Advanced Study and the Kavli Institute for Theoretical Physics. I obtained my Bachelor’s and Master’s degrees in Physics at La Sapienza under the supervision of Giorgio Parisi and Alessandro Treves (SISSA).
Outside of research, I am endlessly curious about the world. I love swimming, playing chess, writing poetry, and building small projects that sit at the intersection of math, games, and human experience.
Recent Publications
Recent papers and preprints.
Combining Induction and Transduction for Abstract Reasoning
Wen-Ding Li, Keya Hu, Carter Larsen, Yuqing Wu, Simon Alford, Caleb Woo, Spencer M. Dunn, Hao Tang, M. Naim, D. Nguyen, Wei-Long Zheng, Z. Tavares, Yewen Pu, Kevin Ellis
Lyfe Agents: Generative agents for low-cost real-time social interactions
Kaiya Zhao, M. Naim, Jovana Kondic, Manuel Ernesto Cortes, Jiaxin Ge, Shuying Luo, Guangyu Robert Yang, Andrew Ahn
Articles
Basis essays and updates this person wrote or contributed to.
Pact: Trustworthy Coordination for Multi-Agentic Ecosystems
Multi-agentic ecosystems are becoming commonplace, but how can we trust them? We propose Pact, a formal coordination language that unifies ideas from game theory, distributed systems and cryptography to enable trustworthy multi-agent coordination.
AutumnBench: World Model Learning in Humans and AI
We’re releasing a new version of Autumn with human baseline results, AI performance comparisons, and an interactive benchmark for world model discovery. This release includes the MARA protocol and provides a public platform for testing causal reasoning capabilities.