Sergio Gutiérrez

My research pursues the long-term objective of autonomous knowledge discovery in the sciences. I conceptualize this as the process of uncovering an environment's dynamics through self-guided interactions, moving beyond traditional supervised learning on curated datasets. My approach will be grounded in experiential training frameworks like Reinforcement Learning (RL), also developing world models that serve as planning and simulation engines. The core interest lies in enabling efficient, task-agnostic, and intrinsically motivated exploration for agents in long-horizon or continual learning settings. At the behavioral level, I want to incentivize agents to prioritize information seeking actions to enhance their autonomy in complex environments.