Zhicheng He

Zhicheng He

he/him

Second-year graduate student in the Department of Biomedical Engineering at the National University of Singapore (NUS).
Research interests: medical image analysis and multi-modal large language models.

Granulon: Awakening Pixel-Level Visual Encoders with Adaptive Multi-Granularity Semantics for MLLM

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

Granulon combines a text-conditioned granularity controller with adaptive visual token aggregation to support fine- and coarse-grained multimodal reasoning.

Recommended citation: Junyuan Mao, Qiankun Li, Linghao Meng, Zhicheng He, Xinliang Zhou, Kun Wang, Yang Liu, Yueming Jin. (2026). "Granulon: Awakening Pixel-Level Visual Encoders with Adaptive Multi-Granularity Semantics for MLLM." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Pages 26317-26327.
Download Paper