Neuroscience
Memory reactivation underlies experience-dependent adaptive regulation of sleep
Experience-specific memory reactivation regulates sleep stability and arousal through hippocampus–amygdala circuits.
Brain · Intelligence · Representation
Guofan Fan 范国藩
I am a Ph.D. student in Pattern Recognition and Intelligent Systems, in a joint program between the Beijing Academy of Artificial Intelligence (BAAI) and the Institute of Automation, Chinese Academy of Sciences (CASIA), advised by Bo Lei and Shan Yu. Previously, I received my M.S. degree in Computer Science, in a joint program between Xi’an Jiaotong University and IIIS, Tsinghua University, advised by Andrew C. Yao and Kaisheng Ma.
My research lies at the intersection of brain science and artificial intelligence, with a focus on brain-inspired AI, 3D computer vision, and multimodal learning.
Research Interests
Brain-Inspired AI
Connecting biological neural mechanisms with artificial intelligence.
3D Computer Vision
Self-supervised representation learning and spatial reasoning for 3D scenes.
Multimodal Learning
Joint representation learning across vision, language, and embodied modalities.
Publications
* Equal contributionFull list on Google Scholar
Embodied AI
SoFar: Language-Grounded Orientation Bridges Spatial Reasoning and Object Manipulation
Language-grounded semantic orientation connects 3D spatial reasoning with fine-grained robotic manipulation.
Brain-Inspired AI
Learning from Pattern Completion: Self-supervised Controllable Generation
A self-supervised framework for controllable image generation, inspired by cortical modularization and hippocampal pattern completion.
Method Overview
3D Computer Vision
Point-GCC: Universal Self-supervised 3D Scene Pre-training via Geometry-Color Contrast
Learning transferable 3D scene representations by connecting geometry and color through self-supervision.
Method Overview
3D Computer Vision
Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining
Generative pre-training guides contrastive learning to bring together complementary 3D representations.
Method Overview
Multimodal Learning
Language-Assisted 3D Feature Learning for Semantic Scene Understanding
Textual scene descriptions guide 3D feature learning toward meaningful geometry and scene context.
Method Overview
Experience
2026.09 – Present
2025.07 – 2026.08
2022.09 – 2025.06
M.S. in Computer Science · Joint Program
Xi’an Jiaotong University & IIIS, Tsinghua University
Advisors: Andrew C. Yao and Kaisheng Ma
2018.09 – 2022.06
B.S. in Biotechnology
Xi’an Jiaotong University