photo
Google Scholar GitHub Twitter LinkedIn Email: mnakamura@umass.edu

About

I’m Mason Nakamura, a third-year Ph.D. student in Computer Science at UMass Amherst, working in the College of Information and Computer Sciences (CICS) as a member of the Resource-Bounded Reasoning Lab, under the guidance of Professor Shlomo Zilberstein. My research is funded by the NSF Graduate Research Fellowship where I broadly work on safety and alignment of AI systems, using methodologies from reinforcment learning and sequential decision-making.

My research interests lie in the intersections of:

  • Value Alignment & AI Safety: Multi-Objective Alignment, Inference-Time Alignment, Preference Learning, Post-Training
  • Multi-Agent Systems: LLM-based Coordination, Multi-Agent VLAs
  • Embodied Agents: Vision-Language-Action Models (VLAs)

Previously, I had internships at the Center for Human-Compatible Artificial Intelligence (CHAI) / Berkeley Artificial Intelligence Research (BAIR) (Summer 2023) on building ethically-compliant autonomous systems using constrained deep reinforcement learning, CHAI/BAIR (Summer-Fall 2022) building a formal framework for computationally bounded agents, Moravian University (Summer 2021 NSF REU) working on extremal and spectral graph theory, and Michigan State University (Summer 2020 NSF REU) working on geometric deep learning.

I’m currently seeking internship opportunities for Summer 2026

Selected Papers

* indicates equal contribution

Terrarium: Revisiting the Blackboard for Multi-Agent Safety, Privacy, and Security Studies
Alignment & Safety Multi-Agent Systems

Mason Nakamura*, Abhinav Kumar*, Saaduddin Mahmud, Sahar Abdelnabi, Shlomo Zilberstein, Eugene Bagdasarian

arXiv 2025 (Under Review)

MAPLE: A Framework for Active Preference Learning Guided by Large Language Models
Alignment & Safety

Saaduddin Mahmud, Mason Nakamura, Shlomo Zilberstein

AAAI 2025

Aligning LLMs on a Budget: Inference-Time Alignment with Heuristic Reward Models
Alignment & Safety

Mason Nakamura*, Saaduddin Mahmud*, Kyle H. Wray, Hamed Zamani, Shlomo Zilberstein

arXiv 2025

Inference-Aware Prompt Optimization for Aligning Black-Box Large Language Models
Alignment & Safety

Saaduddin Mahmud, Mason Nakamura, Kyle H. Wray, Shlomo Zilberstein

arXiv 2025 (Under Review)

BLAST: Bayesian Learning for Adaptive Selection of Transitions for Continual Reinforcement Learning
Continual Learning

Mason Nakamura, Saaduddin Mahmud, Shlomo Zilberstein

AAAI 2025 GenPlan Workshop

Formal Composition of Robotic Systems as Contract Programs
Embodied Agents

Mason Nakamura, Justin Svegliato, Samer Nashed, Shlomo Zilberstein, Stuart Russell

IROS 2023

News

2025

June

2024

December

2023

June

May

  • I graduated summa cum laude from Marist College, majoring in applied mathematics and data science with a minor in computer science.

March

  • I accepted a computer science MS/PhD offer at UMass Amherst in the Resource-Bounded Reasoning Lab directed by Shlomo Zilberstein.

2022

May-November

  • Visiting scholar at the Center for Human-Compatible AI (CHAI) at UC Berkeley on building well-founded AI using metareasoning.

March

  • Our paper on using combinatorics to solve a combinatorial geometry problem has been accepted to MAA Mathematics Magazine!

2021

May-August

  • Visiting scholar at Moravian University as part of an NSF REU on computational methods in discrete mathematics working on spectral and extremal graph theory.

2020

May-August

  • Visiting scholar at Michigan State University as part of an NSF REU on experimental mathematics working on geometric deep learning.