GeoVerse Lab
โ† AI & Computing Sciences Division

Reinforcement Learning & Agents Center

Develops reinforcement learning theory, autonomous agent architectures, and sequential decision-making systems - spanning deep RL, model-based world models, multi-agent systems, and LLM-based agents - for scientific exploration, experiment design, and adaptive control.

Machine Learning CenterNatural Language Processing CenterComputer Vision CenterHigh-Performance Computing CenterQuantum Computing CenterGenerative AI & Foundation Models CenterReinforcement Learning & Agents CenterAI Safety & Alignment CenterAI for Science CenterRobotics & Embodied AI Center
Research Fields 15
Markov Decision Processes & Optimal Stochastic Control
๐Ÿ—บ๏ธ Markov Decision Processes & Optimal Stochastic Control
A Markov decision process is the standard mathematical model of sequential decision-makingโ€ฆ
Temporal-Difference & Associative Learning
๐Ÿ”” Temporal-Difference & Associative Learning
Temporal-difference (TD) learning updates predictions from the difference between successiโ€ฆ
Policy Gradient & Direct Policy Optimization
โ›ฐ๏ธ Policy Gradient & Direct Policy Optimization
Policy gradient methods parameterise the agent's policy directly, as a probability distribโ€ฆ
Actor-Critic Methods & Hedonistic Neuron Theory
๐ŸŽญ Actor-Critic Methods & Hedonistic Neuron Theory
Actor-critic architectures split a reinforcement-learning agent into two cooperating modulโ€ฆ
Model-Based RL & World Models
๐Ÿ”ฎ Model-Based RL & World Models
Model-based reinforcement learning equips an agent with an internal predictive model of itโ€ฆ
Multi-Agent Systems & Stochastic Games
๐Ÿ•ธ๏ธ Multi-Agent Systems & Stochastic Games
Multi-agent systems study decision-making when several learners share one environment, eacโ€ฆ
Monte Carlo Tree Search & Self-Play Game Learning
โ™Ÿ๏ธ Monte Carlo Tree Search & Self-Play Game Learning
Monte Carlo tree search (MCTS) decides moves by growing, one simulation at a time, a searcโ€ฆ
LLM-Based Autonomous Agents & Goal-Directed Behavior Theory
๐Ÿค– LLM-Based Autonomous Agents & Goal-Directed Behavior Theory
LLM-based agents are software systems in which a large language model supplies the reasoniโ€ฆ
Bayesian & Optimal Experimental Design for Science
๐Ÿงช Bayesian & Optimal Experimental Design for Science
Optimal experimental design chooses experimental conditions, what to measure, where, and wโ€ฆ
Exploration-Exploitation & Multi-Armed Bandit Theory
๐ŸŽฐ Exploration-Exploitation & Multi-Armed Bandit Theory
The exploration-exploitation dilemma is the fundamental tension of sequential decision-makโ€ฆ
Hierarchical Reinforcement Learning & Goal Decomposition
๐Ÿชœ Hierarchical Reinforcement Learning & Goal Decomposition
Hierarchical reinforcement learning introduces temporally extended actions into the MDP frโ€ฆ
Imitation Learning & Observational/Social Learning
๐Ÿชž Imitation Learning & Observational/Social Learning
Imitation learning acquires behaviour from demonstrations by a competent agent rather thanโ€ฆ
Intrinsic Motivation & Curiosity-Driven Exploration
โœจ Intrinsic Motivation & Curiosity-Driven Exploration
Intrinsic motivation research builds reward signals that come from inside the learning ageโ€ฆ
Sequential Games & Extensive-Form Analysis
โ™Ÿ๏ธŽ Sequential Games & Extensive-Form Analysis
Extensive-form game theory represents strategic interaction as a game tree: players move iโ€ฆ
Adaptive & Self-Tuning Control Systems
๐ŸŽ›๏ธ Adaptive & Self-Tuning Control Systems
Adaptive control is the branch of control theory in which the controller adjusts its own pโ€ฆ