Yiqi Wang

I am currently a Research Associate at Robotics Institute (RI), Carnegie Mellon University. Previously, I graduated as a master student in robotics (MSR) and ECE, working with Prof. Jeff Schneider, Prof. Yuejie Chi and Prof. Chris Atkeson. I received my BS from University of Wisconsin-Madison in Computer Science, and worked as an undergraduate researcher at Informatics Skunkworks, under the supervision of Prof. Dane Morgan.

Email  /  Resume (Nov, 2025)  /  Google Scholar  /  Github  / 

profile photo


Research Interests

I interested in developing sample-efficient algorithms for robot learning. This includes:

  1. How to learn representations from existing data that are in-the-wild (e.g., internet-scale images, videos), and across embodiments (robots, human)?
  2. How to leverage the learned representations to faciliate sample-efficient robot learning and online improvement (e.g., RL)?
Real-world robot data is small in size and can hardly match the scale of online visual data (in the wild), which contains useful information about common objects and real-world spaces and grows larger on a daily basis. Being capable of learning from this data could lead to prior knowledge with basic "understanding" of common objects/scenes, easily adaptable to a new task or robot. Recently, I got pretty excited about using Self-Supervised Learning (SSL) to digest existing data that are multi-domain and multi-embodiment. The resulting SSL representation will be used to build a downstream robot system (policy, world model, reward function) with a small amount of real-world robot data and help the robot to make reliable decisions via online improvement (RL).



Projects

Latent Policy Steering with Embodiment-Agnostic Pretrained World Models
Yiqi Wang, Mrinal Verghese, Jeff Schneider,
Under Review, 2026
[paper] [presentation] [website]
[read more]

Scalable Dynamic Resource Allocation via Domain Randomized Reinforcement Learning
Yiqi Wang, Laixi Shi, Martin Hyungwoo Lee, Jaroslaw Sydir, Zhu Zhou, Yuejie Chi, Bin Li,
IEEE GLOBECOM, 2024
[paper] [poster]
[read more]

A trajectory is worth three sentences: multimodal transformer for offline reinforcement learning
Yiqi Wang, Mengdi Xu, Laixi Shi, Yuejie Chi,
Proceedings of the Thirty-Ninth Conference on Uncertainty in Artificial Intelligence, (UAI), 2023
[paper] [code] [presentation]
[read more]

Automatic Speech Recognition Meets Language Modeling
Yiqi Wang, Jianyu Mao, Aditya Rathod, [code]
[read more]

Website template from Jon Barron.