KUN LI RESEARCH GROUP / TSINGHUA AIR / RSI SCIENCE

Building Self-Improving Scientific Intelligence.

We combine scientific agents, world models, and high-performance computing to enable verifiable scientific discovery and recursively improve the models, methods, and tools behind it.

Explore research Meet the team

Research in Motion

Research Highlights.

Four representative projects, visualized through restrained scientific line art.

Swipe to explore · Tap for details

Open arXiv

SCIENTIFIC WORLD MODELS

AtomWorld-Mirror

Research highlight / 01

AtomWorld-Mirror

Macro-step world modeling of long-horizon materials evolution.

SCIENTIFIC WORLD MODELS

Open arXiv
Open arXiv

AI-DRIVEN SIMULATION

SwarmThinkers

Research highlight / 02

SwarmThinkers

Physically consistent, learned atomic KMC transitions at scale.

AI-DRIVEN SIMULATION

Open arXiv
Open arXiv

ACCELERATOR-NATIVE QUANTUM CHEMISTRY

MakoXC

Research highlight / 03

MakoXC

Matrix-aligned exchange–correlation evaluation for DFT.

ACCELERATOR-NATIVE QUANTUM CHEMISTRY

Open arXiv
Open paper

SCALABLE SCIENTIFIC COMPUTING

FlashFFTStencil

Research highlight / 04

FlashFFTStencil

FFT-based dense stencil computation on tensor cores.

SCALABLE SCIENTIFIC COMPUTING

Open paper

Illustrative diagrams based on the research concepts; not original simulation trajectories.