Empowerment

Follow this topic Ask about this topic

Latest papers

10 papers5 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

Measuring Empowerment from Offline Data for Intrinsically Motivated Exploration
Oct 17, 2026Ishir Garg, Vivek Myers, Qiyang Li +3Empowerment RL Exploration
UC Berkeley · Princeton University
Scaling Empowered Agents
Oct 9, 2026Catherine Ji, Vivek Myers, Joshua Lin +3Empowerment
Department of Physics Princeton University · Department of Computer Science UC Berkeley · Department of Mathematics Princeton University +1
The Geometry of Empowerment
Oct 6, 2026Catherine Ji, Vivek Myers, Sergey Levine +1Empowerment
Department of Computer Science UC Berkeley · Department of Computer Science Princeton University
A Unified Framework for Empowerment and Predictive Control
Oct 2, 2026Wooyoung Chung, Tristan Shah, Volodymyr Makarenko +1Empowerment Model Predictive Control
Department of Computer Science, Whitacre College of Engineering, Texas Tech University.
PowerBench: Measuring Language Model Bias in Power-shifting Requests
Oct 1, 2026Nicolas Martorell, Wendy Brau, Gonzalo A. Heredia +3Language Model Bias Evaluation Refusal Behavior in Language Models
Universidad de Buenos Aires · Buenos Aires AI Safety Hub · The World Bank +1
Learning to Perceive the World Through Control: Empowerment-Based Representation Learning
May 28, 2026Mahsa Bastankhah, Sophie Broderick, Benjamin EysenbachRepresentation Learning Reinforcement Learning
Princeton University, USA
Learning to Choose: An Empowerment-Guided Multi-Agent System with semantic communication for Adaptive Method Selection
May 28, 2026Geremy Loachamín-Suntaxi, Robert Lazar, Dimitrios G. Giovanis +2Multi-Agent LLM Systems AI Agents for Scientific Discovery
School of Chemical Engineering, National Technical University of Athens, Zographos Campus, 15780, Attiki, Greece · Luxembourg Institute of Science and Technology Esch-sur-Alzette, L-4362, Luxembourg
Prediction and Empowerment: A Theory of Agency through Bridge Interfaces
May 7, 2026Richard CsakyPartially Observable Markov Decision Processes Empowerment
Active Inference: A method for Phenotyping Agency in AI systems?
Apr 25, 2026Philip Wilson, Axel Constant, Mahault Albarracin +4Active Inference AI Governance
Independent Researcher · Department of Engineering and Informatics, University of Sussex, Falmer, Brighton, BN1 9RH, UK · Laboratoire d’Analyse Cognitive de l’Information, Université du Québec à Montréal, Québec, Canada +4
Multi-Agent Empowerment and Emergence of Complex Behavior in Groups
Apr 22, 2026Tristan Shah, Ilya Nemenman, Daniel Polani +1Multi-Agent Coordination Intrinsic Motivation
Department of Computer Science Texas Tech University United States of America · Department of Physics, Department of Biology, Initiative in Theory and Modeling of Living Systems Emory University United States of America · Department of Computer Science, School of Physics, Engineering and Computer Science University of Hertfordshire United Kingdom

Oct 17, 2026J/K move · Enter open · S save

Measuring Empowerment from Offline Data for Intrinsically Motivated Exploration

Ishir Garg, Vivek Myers, Qiyang Li, Catherine Ji+2

UC Berkeley · Princeton University

Empowerment RL Exploration Intrinsic Motivation Offline-To-Online RL

Open paper AskPDF arXiv

Large, unlabeled datasets of prior experience have become increasingly available for agentic tasks in domains such as robotics and coding. A common recipe for these tasks is to train a policy using the prior data, and then further improve it with online rollouts. But while much work has studied how to use both the offline and online data for policy extraction, far less work has been done to understand how the prior data can be used to efficiently explore during the online learning phase. Previous work has attempted to formalize this idea by defining the empowerment, a measure of how many future states an agent can realize; states with higher empowerment require more exploration as there are more futures to explore. However, existing methods for estimating empowerment require online policy rollouts in the environment. We propose the first method for estimating empowerment from unlabeled prior data, without any environment interaction. By jointly learning an empowerment-maximizing policy together with its discounted state-occupancy measure, our estimator can be trained on unlabeled trajectories generated from any unknown policy. We demonstrate that our empowerment estimator qualitatively gives intuitive empowerment measures and show that the pretrained estimator can be used to improve exploration in the online goal-conditioned reinforcement learning setting. Code is available at https://github.com/ishirgarg/offline-empowerment.