Steven Zhang

Steven Zhang

he/him/his

Second-Year Master's Student at CMU

Publications

You can also find my articles on my Google Scholar profile.

Bridging the EHR Divide: Asymmetric Contrastive Learning for Cross-National Medical Representation Transfer

Published in NeurIPS 2026 GenAI4Health Workshop (Accepted, Poster), 2026

This paper introduces Asymmetric SupCon, a task-specific contrastive pre-training objective for cross-national EHR representation transfer. Pre-trained on longitudinal records from 3.98 million Taiwanese NHIRD patients, the models improve over random initialization on MIMIC-IV and demonstrate strong few-shot disease-prediction performance on EHRSHOT.

Download Paper

BURExtract-Llama: An LLM for Clinical Concept Extraction in Breast Ultrasound Reports

Published in MCHM ’24: Proceedings of the 1st International Workshop on Multimedia Computing for Health and Medicine, 2024

BURExtract-Llama extracts structured clinical information from breast ultrasound reports using a Llama 3–8B model fine-tuned on GPT-4-generated annotations. It achieves an average F1 score of 84.6% on clinician-annotated reports, comparable to GPT-4, while supporting lower-cost, in-house deployment.

Download Paper