Publication
2026
PaCX-MAE: Physiology-Augmented Chest X-Ray Masked Autoencoder
ICML 3rd Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences, 2026
Developed a cross-modal distillation framework that incorporates physiological information from paired ECG and laboratory data into chest X-ray representations during pretraining, requiring only chest X-rays at inference. Evaluation across nine benchmarks shows improved performance over domain-specific masked autoencoding, particularly on physiology-dependent tasks and with limited labeled data, while maintaining segmentation performance.
2025
Evaluating Graphical Perception with Multimodal LLMs
IEEE Pacific Visualization Conference, 2025
Evaluated pretrained and fine-tuned multimodal large language models on graphical-perception benchmarks, comparing model accuracy against established human baselines. Comparing pretrained and fine-tuned models with human performance reveals that the models outperform humans on some tasks but fall short on others.
Work in Progress
