Saved in:
Bibliographic Details
Main Authors: Richens, Jonathan, Abel, David, Bellot, Alexis, Everitt, Tom
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.01622
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911218843779072
author Richens, Jonathan
Abel, David
Bellot, Alexis
Everitt, Tom
author_facet Richens, Jonathan
Abel, David
Bellot, Alexis
Everitt, Tom
contents Are world models a necessary ingredient for flexible, goal-directed behaviour, or is model-free learning sufficient? We provide a formal answer to this question, showing that any agent capable of generalizing to multi-step goal-directed tasks must have learned a predictive model of its environment. We show that this model can be extracted from the agent's policy, and that increasing the agents performance or the complexity of the goals it can achieve requires learning increasingly accurate world models. This has a number of consequences: from developing safe and general agents, to bounding agent capabilities in complex environments, and providing new algorithms for eliciting world models from agents.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01622
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle General agents contain world models
Richens, Jonathan
Abel, David
Bellot, Alexis
Everitt, Tom
Artificial Intelligence
Machine Learning
Robotics
Are world models a necessary ingredient for flexible, goal-directed behaviour, or is model-free learning sufficient? We provide a formal answer to this question, showing that any agent capable of generalizing to multi-step goal-directed tasks must have learned a predictive model of its environment. We show that this model can be extracted from the agent's policy, and that increasing the agents performance or the complexity of the goals it can achieve requires learning increasingly accurate world models. This has a number of consequences: from developing safe and general agents, to bounding agent capabilities in complex environments, and providing new algorithms for eliciting world models from agents.
title General agents contain world models
topic Artificial Intelligence
Machine Learning
Robotics
url https://arxiv.org/abs/2506.01622