The Geometry of Empowerment

cs.LG, cs.AI

Submitted: 2026-10-06

Updated: 2026-10-07

Comments: 34 pages, 12 figures

Project page: https://empowerment-geometry.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Information-Theoretic Policy Pre-Training with Empowerment
• Optimal Policies Tend to Seek Power
• Episodic Novelty Through Temporal Distance
• Offline Goal-conditioned Reinforcement Learning with Quasimetric Representations
• Dual Goal Representations
• Reward is not Necessary: How to Create a Modular & Compositional Self-Preserving Agent for Life-Long Learning
• When Assisting One Disempowers Another
• Which Features are Best for Successor Features?
• Control Capacity of Partially Observable Dynamic Systems in Continuous Time
• Clarabel: An interior-point solver for conic programs with quadratic objectives
• Efficient Empowerment Estimation for Unsupervised Stabilization

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks