🤖 Agent Lab

@ University of Liverpool

We are a research group dedicated to building AI agents that can reason, learn from interaction, and collaborate with others.

Our research focuses on Reasoning, Reinforcement Learning, and Multi-Agent Learning.

Research

🧠 Reasoning

Building agents that can reason, plan, solve complex problems, and improve their capabilities through experience and self-evolution.

LLM Agents Planning Reasoning Self-Evolution

🎯 Reinforcement Learning

Developing agents that learn through interaction, with an emphasis on exploration, generalization, continual learning, and safe decision-making.

Exploration Generalization Continual Learning Safe RL

🤝 Multi-Agent Learning

Studying how multiple agents cooperate, coordinate, communicate, and develop social intelligence in shared environments.

Cooperation Coordination MARL Social Intelligence

Members

Meng Fang Director
Botond Hamori Student
Yuxuan Huang Student
Xiao Xiao Student
Ke Xu Student
Harry Mackenzie Student

Featured Projects

We explore Agent + X across a range of domains and environments.

Agentic Web
Building agents that can reason, act, and improve in open-ended web environments.
Q-Evolve Agentic Web Survey WebDecept
Math Reasoning
Benchmarking and improving mathematical problem-solving capabilities in AI agents.
MathOdyssey
Text Adventure Games
Language-based agents for reasoning, planning, exploration, and long-horizon decision-making.
Monte Carlo Planning Stay Moral and Explore
Vision-and-Language RL
Learning agents that perceive, reason, and act from visual and language observations and feedback.
HASARD Geo-R
Continual Learning
Developing agents that continually learn, adapt, and improve their capabilities over time.
COOM MEAL

Selected Publications (Full List)

Join Agent Lab

We are looking for motivated PhD students, interns, and visiting researchers interested in building the next generation of AI agents. Get in touch →