INSPIRE Lab Imaging- and Neuro-computations for Precision Informatics Research

I’m Junhao Zhao, a Ph.D. student in Electrical and Computer Engineering at the University of Maryland, College Park, advised by Nan Xu. I received bachelor’s degrees in Mathematics and Physics and in Mechanical Engineering from Tsinghua University, where I was fortunate to work as a research intern in Prof. Gao Huang’s group. My research asks how autoregressive models use history to make predictions, and when that history provides reliable learning signals. My recent work explores credit assignment in long-horizon forecasting and the role of state-transition complexity in brain activity prediction. I’m interested in extending these insights to reinforcement learning, world models, and long-horizon LLM agents. I’m seeking research internship opportunities for Summer 2027.

Publications

* denotes equal contribution † denotes equal supervision

Large Distant Gradients Need Not Be Reliable: Reliability-weighted Credit Assignment for Long-horizon Autoregressive Forecasting

Junhao Zhao, David Michael Simberg, Jacob Kang, Colin Connor Kurniawan, Nan Xu arXiv preprint, 2026 [Link]

CARE: Contrastive Alignment for ADL Recognition from Event-Triggered Sensor Streams

Junhao Zhao, Zishuai Liu, Ruili Fang, Jin Lu, Linghan Zhang, Fei Dou IEEE International Conference on Pervasive Computing and Communications (PerCom), 2026 [Link] [Code]

Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment

Jiayi Guo, Junhao Zhao, Chaoqun Du, Yulin Wang, Chunjiang Ge, Zanlin Ni, Shiji Song, Humphrey Shi, Gao Huang IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025 [Link] [Code]

Search for Junhao (Jay) Zhao's papers on the Publications Page