Open-source adoption
GitHub6,336GitHub Stars
4 public repositories · AIMIA-PKU ↗ · RT-DETR ↗
Cumulative star events · weekly
Updated
PEKING UNIVERSITY · SHENZHEN
AIMIA Lab
AIMIA 实验室
Our research focuses on embodied intelligence, multimodal intelligence and AI for Science (AI4S).
GROUP LEADJie Chen 陈杰
Associate professor
School of Electronic and Computer Engineering
Peking University Shenzhen Graduate School
Research in numbers
Academic citations
Google Scholar30,083Total citations
Jie Chen · h-index 64
Cumulative citations · since 2010
Updated
Research at a glance
Hierarchical planning and action
An agent planner decomposes goals, while a native-action agent or VLA model executes actions. Environment feedback informs the next plan.
Explore the process
Research areaSelect an action to explore planning and execution.
Connecting vision and language
Studying the relationships between images, video and language through cross-modal representations and alignment.
Explore the process
Research areaExplore the alignment between scene regions and language concepts.
From representations to prediction
Combining data and learning methods to study biomolecular representations, property prediction and scientific modeling.
Explore the process
Research areaIllustration: build molecular representations from an amino-acid sequence.
Interactive research illustrations
Latest publications
All outputsSix latest publications · by publication year
- 2026TPAMIUnified Granularity Controller for Interactive Segmentation
- 2026TPAMIAdversarial Purification by Consistency-aware Latent Space Optimization on Data Manifolds
- 2026Nature CommunicationsSplitSeek-Pro: accurate prediction of splittable sites on protein structures
- 2026Medical Image AnalysisSwitch-UMamba: Dynamic scanning vision Mamba UNet for medical image segmentation
- 2026TCSVTDemoChain: Decoupled Motion-Chain for Multi-Action Video Generation
- 2026IJCVKnowing Where to Focus: Attention-Guided Alignment for Text-based Person Search
Honors & awards
All honorsAwards & academic recognition
- 2025
First Prize, National Science and Technology Progress Award.
- 2025
Zhiwei Nie received a Youth Outstanding Paper Award at the World Artificial Intelligence Conference for “A unified evolution-driven deep learning framework for virus variation driver prediction”.
- 2024
Second Prize in the national final and First Prize in Guangdong, 2024 “Data Elements ×” Competition.
- 2024
“Zu Chongzhi” Annual Major Achievement Award for Frontier Innovation in Artificial Intelligence.
- 2023
Special Prize, Guangdong Science and Technology Progress Award.
- 2022
The joint team involving our group was one of three finalists worldwide for the ACM Gordon Bell Special Prize for HPC-Based COVID-19 Research, and the only finalist team from China.
- 2022
First place on a CAMEO monthly leaderboard for protein structure prediction.
RESEARCH
Explore our research Research Outputs
Embodied intelligence
How can an agent planner collaborate with native-action agents or VLA models to act in the physical world?
Exploring a hierarchical architecture with an agent planner and either a native-action agent or a VLA model, connecting task planning, action execution and environment feedback.
Research enquiriesMultimodal intelligence
How can models understand and generate content across visual and language modalities?
Understanding and generating content across vision and language, including images, videos and human motion.
Related publication · TCSVT 2026DemoChain: Decoupled Motion-Chain for Multi-Action Video GenerationAI for Science (AI4S)
How can AI support scientific prediction and discovery in biology and medicine?
Developing AI methods for biomolecular prediction and design, virus evolution, protein structure and medical image analysis.
Related publication · Nature Communications 2026SplitSeek-Pro: accurate prediction of splittable sites on protein structuresRESEARCHERS & STUDENTS
Team & Culture
Our group brings together doctoral researchers and master's students working across our research areas.
Meet the group, explore members' research interests and find our alumni directory.
FOR PROSPECTIVE STUDENTS
Graduate study & research
Students interested in embodied intelligence, multimodal intelligence or AI for Science (AI4S) can find contact details and admissions information here.
Prospective studentsAIMIA · AROUND THE WORLD
Visitor footprint
A shared curiosity, across places and time zones.
Pageviews & geographic distributionApproximate locationsMapMyVisitors ↗