William Pinstrup Huey

Hi, I’m Will. I’m a Visiting Researcher at the University of Washington, where I’m working on dexterous manipulation with Prof. Abhishek Gupta. Previously, I worked with Prof. Sanjiban Choudhury on learning robot control from videos. My research focuses on learning control priors from off-domain data and adapting them to the real world.

I recently graduated Summa Cum Laude from Cornell with a B.S. in Computer Science (‘25). I maintained a 4.1 GPA, I was a teaching assistant for Visual Imaging in the Electronic Age, and I was a Rawlings Presidential Research Scholar. I spent my summers interning at NASA, the Robotic Systems Lab, and Amazon Web Services.

I also enjoy climbing, trail running, and watching soccer (coyg).

News

Jul 01, 2025	I started as a Visiting Researcher with Prof. Abhishek Gupta at the University of Washington
May 22, 2025	I graduated Summa Cum Laude from Cornell with a B.S. in Computer Science
May 01, 2025	Imitation Learning from a Single Temporally Misaligned Video was accepted to ICML 2025
Mar 04, 2025	I won $500 and first place at the RCPRS Grand Slam for my presentation on learning from videos
Feb 24, 2025	I was a panelist at ACSU research night, where I had the opportunity to share my awesome experience doing research at Cornell with other undergraduates

Selected Publications

Imitation Learning from a Single Temporally Misaligned Video

William Huey^*, Huaxiaoyue Wang^*, Anne Wu, Yoav Artzi, and Sanjiban Choudhury

In International Conference on Machine Learning (ICML), 2025

TLDR arXiv Code

To learn from a single video, current approaches leverage an image encoder, distance metric on the encoder’s latent space, and distribution-matching algorithm to produce dense rewards for RL. We prove that frame-level distribution matching algorithms cannot accurately model the ground truth reward of a video demonstration. We propose ORCA (ORdered Coverage Alignment), an intuitive algorithm that matches distributions at the sequence level, achieving better performance on 16 video-imitation tasks in two simulation environments.