Knowledge-integrated Machine Intelligence
KiMI is dedicated to Knowledge-integrated Machine Intelligence, building AI systems that can understand, retrieve, reason, and act across modalities and domains by integrating data-driven learning, domain knowledge, and foundation models.
Our broader goal is to make these systems more reliable, efficient, and impactful for real-world scientific and engineering problems.
Research Directions
Four converging tracks where the lab invests its long-horizon research effort.
Multimodal Reasoning
Understanding and reasoning across vision, language, and structured knowledge.
- Vision-Language
- Structured Knowledge
- Reasoning Eval
RAG Systems
Multimodal search, retrieval, ranking, and grounded generation over large-scale corpora.
- Search
- Ranking
- Grounding
LLM Agents
Planning, memory, tool use, verification, and multi-agent collaboration for complex problem solving.
- Planning
- Tool Use
- Multi-Agent
AI for X
Knowledge-enhanced foundation models and intelligent systems for biology, materials, and engineering.
- Biology
- Materials
- Engineering
Main Collaborators
Senior researchers with whom the group sustains long-term joint projects and student supervision.
Students and Researchers
Current group members, grouped by training stage and listed in order of joining.
Ph.D.
7Zhuangfei Hu
- Efficient Large Models
Siyi Liu
- Data Synthesis
- Multimodal Retrieval
Hengwei Dai
- GraphRAG
- LLM for Code
Yuhang He
- Reinforcement Learning
- LLM Infra
Xiaotang Wang
- Graph Learning
- AI4Science
Hange Zhou
- Agentic Reasoning
- Embodied AI
Haodong Wu
- Reasoning RL
- LLM Agents
MPhil
6Runqing Xu
- LLM Agents
- AI for X
Mingkai Qiu
- Drug Discovery
Hongyu Ge
- Reasoning RL
- Agentic Reasoning
Xiaorong Zhu
- Multimodal Retrieval
- Game Agents
Boyang Zhao
- Embodied Memory
- Vision-Language Navigation
Bairui Zhang
- Agentic Reasoning
RA
3Enjun Du
- KG Reasoning
- Multimodal RAG
Yijia Li
- State Space Model
Zirong Chen
- Multimodal Retrieval