Hi, I'm Zhonghao Shi
I'm a Postdoc Fellow at
Harvard University
Harvard University
My research focuses on building machines that can truly understand humans. I am passionate about research topics including human user simulation, benchmarking and post-training of multimodal (audio-visual) language models, and human-AI interaction in application domains such as education.
I am currently a postdoctoral fellow working with Prof. Ying Xu at Harvard University. I defended my PhD at the University of Southern California (USC) where I was fortunate to be advised by Prof. Maja Matarić. Previously I worked on trustworthy machine learning at JPMorganChase and studied at University College London (UCL).
Featured Publications
My latest research on human-centered AI and machine learning.
We used two validated games from the cognitive science literature to systematically study how well recent LLMs predict player actions and whether they can leverage and generalize players' underlying motives.
ACL Findings
2026
We introduced ChildVox, a benchmark for characterizing the diverse acoustic signals through which children communicate, following the full developmental trajectory from birth through school age.
arXiv
2026
We proposed CMA-ES-IG, an algorithm that explicitly incorporates user experience considerations into the preference learning process by suggesting perceptually distinct and informative trajectories for users to rank.
arXiv
2026
We investigated the effectiveness of test-time adaptation for child speech recognition, with the goal of enabling continuous, unsupervised adaptation at test time.
Interpseech
2025
We introduced HRIBench, a visual question-answering (VQA) benchmark designed to evaluate VLMs across a diverse set of human perceptual tasks critical for HRI.
ISER
2025
We applied supervised machine-learning algorithms to model user engagement in the context of long-term, in-home SAR interventions for children on the autism spectrum.
Science Robotics
2020