Hyun Joe Jeong

I am a first year Robotics Institute PhD student at Carnegie Mellon University, advised by Andrea Bajcsy. Previously, I graduated from UC San Diego, where I was advised by Sylvia Herbert. I have also collaborated with Somil Bansal. I am fortunate to be a NSF GRFP Fellow.

Email  /  Scholar  /  Twitter  /  Github

Research

My research focuses on developing methods that learn and leverage task-relevant representations for efficient control and decision-making. I draw on ideas from generative modeling, representation learning, and reinforcement learning to build agents that can reason, adapt, and generalize across complex environments. Representative papers are highlighted.

We learn a language feedback policy that steers a frozen VLA at test time—deciding what to say and, crucially, when not to say it—with conformal guarantees against harmful steering.

We demonstrate that RBNs are highly accurate in low dimensions, and safer than the standard neural CBF approach in high dimensions.

We enable robots to know when they can confidently execute a user's desired goal, and automatically suggest safe alternatives when they cannot.

Parameterized Fast and Safe Tracking can smoothly trade off between the navigation speed and the tracking error (therefore maneuverability) while guaranteeing obstacle avoidance in a priori unknown environments.

we propose a method to decompose systems of a particular coupled nonlinear structure, in order to solve for the CLVF in each low-dimensional subsystem. We then reconstruct the full-dimensional CLVF and provide sufficient conditions for when this reconstruction is exact.

Template adapted from here.