Marco Cusumano-Towner

I am currently a research scientist at Apple working with Vladlen Koltun on autonomous systems and physical intelligence.

I completed my PhD in EECS at MIT, where I was advised by Vikash Mansinghka and Josh Tenenbaum. My PhD research focused on compositional abstractions for generative probabilistic models and inference algorithms operating on those models. During my PhD I created the Gen probabilistic programming system. Prior to MIT, I was an early employee and technical lead at an early-stage molecular diagnostics startup backed by Sequoia Capital. I completed my MS in computer science at Stanford. I completed my BS in EECS at UC Berkeley, where I worked with Pieter Abbeel on household robotics. My academic research has been funded by the NSF GRFP and the NDSEG fellowship.

Papers

In submission, 2025

Reinforcement Learning for Long-Horizon Interactive LLM Agents

Kevin Chen*, Marco Cusumano-Towner*, Brody Huval*, Aleksei Petrenko*, Jackson Hamburger, Vladlen Koltun, Philipp Krähenbühl

In submission, 2025

In submission, 2025

Robust Autonomy Emerges from Self-Play

Marco Cusumano-Towner*, David Hafner*, Alex Hertzberg*, Brody Huval*, Aleksei Petrenko*, Eugene Vinitsky*, Erik Wijmans*, Taylor Killian, Stuart Bowers, Ozan Sener, Philipp Krähenbühl, Vladlen Koltun

In submission, 2025

Recursive Monte Carlo and Variational Inference with Auxiliary Variables

Alexander K. Lew, Marco Cusumano-Towner, Vikash K. Mansinghka

UAI 2022

Interval Estimators of Entropy and Information Measures via Inference in Probabilistic Models

Feras A. Saad, Marco Cusumano-Towner, Vikash K. Mansinghka

AISTATS 2022

3DP3: 3D Scene Perception via Probabilistic Programming

Nishad Gothoskar, Marco Cusumano-Towner, Ben Zinberg, Matin Ghavamizadeh, Falk Pollok, Austin Garrett, Joshua B. Tenenbaum, Dan Gutfreund, Vikash K. Mansinghka

NeurIPS 2021

Automating Involutive MCMC using Probabilistic and Differentiable Programming

Marco Cusumano-Towner, Alexander K. Lew, Vikash K. Mansinghka

arXiv 2020

Trace types and denotational semantics for sound programmable inference in probabilistic languages

Alexander K. Lew, Marco Cusumano-Towner, Benjamin Sherman, Michael Carbin, Vikash K. Mansinghka

POPL 2020

Gen: a general-purpose probabilistic programming system with programmable inference

Marco Cusumano-Towner, Feras A. Saad, Alexander K. Lew, Vikash K. Mansinghka

PLDI 2019

Bayesian synthesis of probabilistic programs for automatic data modeling

Feras A. Saad, Marco Cusumano-Towner, Ulrich Schaechtle, Martin C. Rinard, Vikash K. Mansinghka

POPL 2019

Structured differentiable models of 3D scenes via generative scene graphs

Ben Zinberg, Marco Cusumano-Towner, Vikash K. Mansinghka

NeurIPS Workshop 2019

Incremental inference for probabilistic programs

Marco Cusumano-Towner, Benjamin Bichsel, Timon Gehr, Martin T. Vechev, Vikash K. Mansinghka

PLDI 2018

Using probabilistic programs as proposals

Marco Cusumano-Towner, Vikash K. Mansinghka

arXiv 2018

Probabilistic programs for inferring the goals of autonomous agents

Marco Cusumano-Towner, Alexey Radul, David Wingate, Vikash K. Mansinghka

arXiv 2017

A social network of hospital acquired infection built from electronic medical record data

Marco Cusumano-Towner, Daniel Y. Li, Shanshan Tuo, Gomathi Krishnan, David M. Maslove

JAMIA 2013

Bringing clothing into desired configurations with limited perception

Marco Cusumano-Towner, Arjun Singh, Stephen Miller, James F. O’Brien, Pieter Abbeel

ICRA 2011

Cloth grasp point detection based on multiple-view geometric cues with application to robotic towel folding

Jeremy Maitin-Shepard, Marco Cusumano-Towner, Jinna Lei, Pieter Abbeel

ICRA 2010