Information-Theoretic Policy Pre-Training with Empowerment

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Schneider, Moritz, Krug, Robert, Vaskevicius, Narunas, Palmieri, Luigi, Volpp, Michael, Boedecker, Joschka
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908579601055744
author Schneider, Moritz
Krug, Robert
Vaskevicius, Narunas
Palmieri, Luigi
Volpp, Michael
Boedecker, Joschka
author_facet Schneider, Moritz
Krug, Robert
Vaskevicius, Narunas
Palmieri, Luigi
Volpp, Michael
Boedecker, Joschka
contents Empowerment, an information-theoretic measure of an agent's potential influence on its environment, has emerged as a powerful intrinsic motivation and exploration framework for reinforcement learning (RL). Besides for unsupervised RL and skill learning algorithms, the specific use of empowerment as a pre-training signal has received limited attention in the literature. We show that empowerment can be used as a pre-training signal for data-efficient downstream task adaptation. For this we extend the traditional notion of empowerment by introducing discounted empowerment, which balances the agent's control over the environment across short- and long-term horizons. Leveraging this formulation, we propose a novel pre-training paradigm that initializes policies to maximize discounted empowerment, enabling agents to acquire a robust understanding of environmental dynamics. We analyze empowerment-based pre-training for various existing RL algorithms and empirically demonstrate its potential as a general-purpose initialization strategy: empowerment-maximizing policies with long horizons are data-efficient and effective, leading to improved adaptability in downstream tasks. Our findings pave the way for future research to scale this framework to high-dimensional and complex tasks, further advancing the field of RL.
format Preprint
id arxiv_https___arxiv_org_abs_2510_05996
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Information-Theoretic Policy Pre-Training with Empowerment
Schneider, Moritz
Krug, Robert
Vaskevicius, Narunas
Palmieri, Luigi
Volpp, Michael
Boedecker, Joschka
Artificial Intelligence
Information Theory
Machine Learning
Robotics
Empowerment, an information-theoretic measure of an agent's potential influence on its environment, has emerged as a powerful intrinsic motivation and exploration framework for reinforcement learning (RL). Besides for unsupervised RL and skill learning algorithms, the specific use of empowerment as a pre-training signal has received limited attention in the literature. We show that empowerment can be used as a pre-training signal for data-efficient downstream task adaptation. For this we extend the traditional notion of empowerment by introducing discounted empowerment, which balances the agent's control over the environment across short- and long-term horizons. Leveraging this formulation, we propose a novel pre-training paradigm that initializes policies to maximize discounted empowerment, enabling agents to acquire a robust understanding of environmental dynamics. We analyze empowerment-based pre-training for various existing RL algorithms and empirically demonstrate its potential as a general-purpose initialization strategy: empowerment-maximizing policies with long horizons are data-efficient and effective, leading to improved adaptability in downstream tasks. Our findings pave the way for future research to scale this framework to high-dimensional and complex tasks, further advancing the field of RL.
title Information-Theoretic Policy Pre-Training with Empowerment
topic Artificial Intelligence
Information Theory
Machine Learning
Robotics
url https://arxiv.org/abs/2510.05996