Test-Time Hinting for Black-Box Vision-Language Models
arXiv preprint · Under review · 2026
PhD Student · UC Berkeley & UCSF
Building trustworthy AI for Healthcare
I am Kaihua (William) Hou, a PhD student in Computational Precision Health at UC Berkeley & UCSF. My work focuses on machine learning methods that are reliable, equitable, and useful in real clinical settings.
News
Recent updates
-
Sep. 2026
I completed my research internship at Alibaba DAMO Academy. Our paper on a generalizable representation learning framework for medical imaging is under review, and a preprint will be available soon.
-
Spring 2026
New pre-print! Test-Time Hinting for Black-Box Vision-Language Models
About
Research with practical impact
I am fortunate to be advised by Ahmed Alaa and Geoff Tison at UC Berkeley & UCSF. Previously, I received my B.S. in Computer Science at Johns Hopkins University, where I was advised by Jithin Yohannan and Mathias Unberath. During my undergraduate studies, I have also had the pleasure to work with Emma Pierson and John Guttag as a research assistant at Massachusetts Institute of Technology.
Focus: robust and equitable machine learning in healthcare, with a strong emphasis on model reliability, representation quality, and AI systems that can reason and assist in clinical workflows.
Roadmap
Path so far
Johns Hopkins University
2019 ~ 2023
UC Berkeley & UCSF
2023 ~ 2028
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
Massachusetts Institute of Technology
Jun. 2022–May 2023
Amazon
May–Aug. 2025
Alibaba
Apr.–Sep. 2026
Publications
Selected work
-
-
Diagnosis by Composition: Detecting and Localizing Unseen Diseases by Composing Learned Radiological Findings
Under review · 2026
-
AI-Enabled Prediction of Incident Heart Failure from Echocardiographic Video
Under review · 2026
-
ReasonEdit: Editing Vision-Language Models using Human Reasoning
ICML · 2026
-
Quantifying disparities in intimate partner violence: a machine learning method to correct for underreporting
npj Women’s Health · 2024
-
Coarse race data conceals disparities in clinical risk score performance
MLHC · 2023
Best Findings Paper (Honorable Mention)
-
Predicting Visual Field Worsening with Longitudinal OCT Data Using a Gated Transformer Network
AAAI / Ophthalmology · 2023
-
A Deep Learning Model Incorporating Spatial and Temporal Information Detects Visual Field Worsening
Scientific Reports · 2023