DIGA@SMU
Distributed Intelligence of Generalist Agents — a research group at the School of Computing and Information Systems, Singapore Management University, led by Xinrun Wang.
About the Group
DIGA mainly focuses on building highly-capable agents — empowered by (multi-agent) reinforcement learning and/or (multimodal) large language models — for complex decision-making tasks. We work from the foundations (algorithms, game-theoretic solvers, world models) up to deployed systems in finance, urban security, and scientific discovery.
Research Directions
Fundamental Decision Making
Single- and multi-agent reinforcement learning, game-theoretic solvers, and unified formulations of decision making.
Decision Making for Applications
RL for FinTech, MARL for urban security, and AI for science — taking algorithms from benchmarks to real deployments.
Foundation Agents
Computer control agents, auto research agents, and the memory/world-model harness that grounds agent behavior.
Computer Science of Language Models
Language models as the new universal computer — and the systems science we still need in order to program them.
Members
Principal Investigator
-
Assistant Professor · Lee Kong Chian Fellow2024.07–
Students & Researchers
-
Shunchao ZhouPhD student2025.08–
-
Ziru ZhouResearch assistant2025.08–
-
Rong HuVisiting student2025.09–
Join Us
Hiring! We are looking for self-motivated people to join the group. Please email me with your background, transcripts, and a short note on which of the directions above you would like to work on.
- PhD / Master students — admission requirements and scholarships are listed on the SMU Computing PhD admissions page.
- Visiting students / CSC-funded researchers — tell me your intended visiting period and research plan.
- Remote interns — we regularly work with remote collaborators on open-source benchmarks and platforms; strong engineering skills are valued.
Group Resources
MatVerse Paper Collection
A regularly updated paper collection on AI for science, particularly materials science.
Computer Science of Language Models
Our running notes on treating language models as a new kind of universal computer.
MATAI Platform
An interactive platform for AI-driven alloy discovery and inverse design.
TradeMaster
An open-source reinforcement learning platform for quantitative trading.
Group publications appear on the publications page.