SurgPace: Complexity-Aware Progressive Distillation for Surgical Multimodal Large Language Models

Published in Medical Image Analysis (accepted), 2026

SurgPace combines teacher-derived sample complexity, an easy-to-hard curriculum, and hierarchical representation alignment to distill compact surgical multimodal large language models. It progressively shifts training from teacher guidance to task supervision and is evaluated on Surg-396K across five dialogue paradigms.

Recommended citation: Zhicheng He, Xiaohan Xing, Hongqiu Wang, Kun Yuan, Guankun Wang, Yanheng Li, Haoxuan Che, Nassir Navab, Hongliang Ren, Long Bai. (2026). "SurgPace: Complexity-Aware Progressive Distillation for Surgical Multimodal Large Language Models." Medical Image Analysis (accepted).
Download Paper