I am a senior in Computer Science at Fudan University focused on medical AI, especially multimodal vision-language models. My work centers on data-centric medical VQA, multimodal model development, and trustworthy step-by-step reasoning for clinically useful AI.
Representative research and engineering work.
[ICLR 2026 Poster]A medical document-to-VQA synthesis pipeline designed for high-quality, verifiable multimodal supervision.
Highlighted papers with links to paper, project resources, and code when available.
Xiaoke Huang*, Ningsen Wang*, Hui Liu, Xianfeng Tang, Yuyin Zhou
ICLR
Generator-verifier LMMs for high-quality medical VQA synthesis with verifiable multimodal supervision.
Recent milestones from papers, preprints, and research activity.
Jan 2026
MedVLSynther was accepted by ICLR 2026.
Read moreOct 2025
Published a preprint on multimodal verifiable question answering synthesis for RL-style training.
Read moreJul 2025
Started as a remote research intern at UCSC VLAA.
Read moreShort notes, workflows, and technical reflections in Markdown/MDX.
How many experiences can avoid me taking bitter lesson again and again?
Several observations made in my research for medical VLM RL.