George Konidaris

From Ticks to Flows: Dynamics of Neural Reinforcement Learning in Continuous Environments

Hierarchical Entity-centric Reinforcement Learning with Factored Subgoal Diffusion

SkillWrapper: Generative Predicate Invention for Skill Abstraction

From Pixels to Factors: Learning Independently Controllable State Variables for Reinforcement Learning

Memory as State Abstraction Over History

Informing Reinforcement Learning Agents by Grounding Language to Markov Decision Processes

Creative Robot Tool Use by Counterfactual Reasoning

Generalizing Subgoals from Single Instances using Hypothesis-Preserving Ensembles

Spectral Collapse Drives Loss of Plasticity in Deep Continual Learning

Skill-Driven Neurosymbolic State Abstractions

Geometry of Neural Reinforcement Learning in Continuous State and Action Spaces

Learning Parameterized Skills from Demonstrations

Knowledge Retention in Continual Model-Based Reinforcement Learning

Knowledge Retention in Continual Model-Based Reinforcement Learning

Informing Reinforcement Learning Agents by Grounding Language to Markov Decision Processes

Learning Large Skillsets in Stochastic Settings with Empowerment

Latent-Predictive Empowerment: Measuring Empowerment without a Simulator

Learning Transferable Sub-goals by Hypothesizing Generalizing Features