Richard Hoffmann

rhoffman@caltech.edu

Hey! I’m Richard, a fourth-year undergrad at Caltech studying Computer Science. I’m advised by Prof. Adam Wierman.

My research interests cover a mix of reinforcement learning, control-theory, generative modeling, and spatial intelligence. I’m excited about multi-agent RL, specifically how intelligence emerges from interacting populations. I’m also interested in generalizing robot learning and planning via video world models, where I work with Yilun Du’s Embodied Minds group.

Before, I worked on post-training with Prof. Tony Yue Yu at Caltech, and before that on predictive vehicle dynamics under Dr. Alec Reed at CU Boulder’s Autonomous Robotics Lab. I’ve interned at Amazon AWS in Seattle and Commerzbank in New York City.

News

Feb 18, 2026	We introduce GMFS, a scalable framework for multi-agent reinforcement learning that maintains near-optimal performance in heterogeneous populations.
Jun 01, 2025	We demonstrate both theoretically and empirically that Practical Adversarial Attacks on Stochastic Bandits via Fake Data Injection can effectively mislead popular algorithms like UCB and Thompson Sampling with minimal attack cost.
Mar 01, 2025	We explore a novel neural population code method to accurately estimate object orientation in Object-Pose Estimation With Neural Population Codes.

Selected Works

Graphon Mean-Field Subsampling for Cooperative Heterogeneous Multi-Agent Reinforcement Learning

Emile Timothy Anand, Richard Hoffmann, Sarah Liaw, and 1 more author

In The Eighteenth Workshop on Adaptive and Learning Agents, 2026

arXiv Code
Learning to Coordinate Symbolic Tools: LLM Agents for Verified Sum-of-Squares Certificates

Bohan Chen, Shivam N. Patel, Richard Hoffmann, and 2 more authors

2026

arXiv
Practical Adversarial Attacks on Stochastic Bandits via Fake Data Injection

Qirun Zeng, Eric He, Richard Hoffmann, and 2 more authors

arXiv preprint, 2025

arXiv Code