mlsanityData to 7 Oct 2026

Learning to Perceive the World Through Control: Empowerment-Based Representation Learning

arXiv 2605.30656 ↗First posted 28 May 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

In many practical reinforcement learning environments, observations are far higher-dimensional than the variables that matter for control. In this work, we ask: can we learn representations that capture only control-relevant features of the environment? We study this question through the empowerment objective, which maximizes an agent's influence over the environment and is widely used for unsupervised skill learning. We show that empowerment agents induce two distinct representations -- forward and backward -- that capture complementary aspects of the state, and both of which are invariant to control-irrelevant features. Thus, empowerment maximization leads agents to learn an implicit, control-centric model of the world. Our analysis highlights the importance of learning representations through interaction rather than from passive datasets: interaction aimed at maximizing control is essential for learning useful invariance properties, a perspective that aligns closely with the causal learning literature.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 24 of its 41 references that are in the map. First posted on arXiv 28 May 2026.

Builds on

24 of 41 references.

References · 24

  1. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,327
  2. 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 950
  3. 2016Deep Variational Information Bottleneckcited by 915
  4. 2018Diversity is All You Need: Learning Skills without a Reward Functioncited by 734
  5. 2020Learning Invariant Representations for Reinforcement Learning without Reconstructioncited by 424
  6. 2016Successor Features for Transfer in Reinforcement Learningcited by 331
  7. 2019Dynamics-Aware Unsupervised Discovery of Skillscited by 308
  8. 2016Variational Intrinsic Controlcited by 262
  9. 2019Provably efficient RL with Rich Observations via Latent State Decodingcited by 169
  10. 2015Variational Information Maximisation for Intrinsically Motivated Reinforcement Learningcited by 142
  11. 2019Fast Task Inference with Variational Intrinsic Successor Featurescited by 131
  12. 2018Variational Option Discovery Algorithmscited by 121
12 more
  1. 2021Learning One Representation to Optimize All Rewardscited by 110
  2. 2020Invariant Causal Prediction for Block MDPscited by 104
  3. 2023METRA: Scalable Unsupervised RL with Metric-Aware Abstractioncited by 80
  4. 2022Denoised MDPs: Learning World Models Better Than the World Itselfcited by 66
  5. 2022Causal Dynamics Learning for Task-Independent State Abstractioncited by 64
  6. 2022Lipschitz-constrained Unsupervised Skill Discoverycited by 60
  7. 2018Learning Actionable Representations with Goal-Conditioned Policiescited by 53
  8. 2022Guaranteed Discovery of Control-Endogenous Latent States with Multi-Step Inverse Modelscited by 42
  9. 2022Bisimulation Makes Analogies in Goal-Conditioned Reinforcement Learningcited by 38
  10. 2021The Information Geometry of Unsupervised Reinforcement Learningcited by 36
  11. 2024Can a MISL Fly? Analysis and Ingredients for Mutual Information Skill Learningcited by 18
  12. 2025Skill Learning via Policy Diversity Yields Identifiable Representations for Reinforcement Learningcited by 4

Built on by

Not cited yet.