Scaling Empowered Agents
What it is
This paper introduces a scalable recipe for maximizing agent empowerment by connecting it to skill learning. It defines empowerment as the number of executable skills from a state and proposes a method to reach states that serve as good starting points for many skills.
The problem it solves
Prior work on empowerment was limited to tabular settings, making it difficult to study empowerment at scale or observe its emergent properties in complex environments.
What you could use it for
- Training agents to exhibit emergent behaviors like resource gathering and survival
- Enabling agents to perform complex tasks such as building structures in Crafter
- Studying the theoretical properties of empowerment in large-scale environments
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.