Shenzhen Research Institute of Big Data

A family of open medical multimodal LLMs (7B and 34B) trained on PubMedVision, a 1.3M-sample medical VQA dataset distilled from PubMed image-text pairs.

Prompt-based medical vision-language pretraining that unifies fusion-encoder and dual-encoder architectures, handling image-only, text-only, and image-text inputs in one model.

Knowledge-enhanced medical vision-and-language pre-training framework that aligns, reasons over, and learns from structured medical knowledge for radiology image-text tasks.

Self-supervised medical vision-and-language pretraining via multi-modal masked autoencoders that reconstruct masked image patches and text tokens.