Expected Free Energy

Latest papers

10 papers2 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

Knowing When to Stop Thinking: A Dissection of Expected Free Energy as a Test-Time Compute Controller for Latent World-Model Planning
Oct 20, 2026Tong Zhang, Zhiyuan Shi, Yun Peng +1Expected Free Energy Inference-Time Compute Allocation
Fudan University · Peking University
Belief-Based Maximum Occupancy Principle and Active Inference
Sep 30, 2026Manolis Mylonas, Rubén Moreno BoteBelief-Space Planning Intrinsic Reward Methods
Center for Brain and Cognition & Department of Engineering Universitat Pompeu Fabra Barcelona, Spain
Expected free energy as an information constraint on the Bethe Lagrangian
Aug 17, 2026Wouter M. KouwActive Inference Expected Free Energy
TU Eindhoven, Eindhoven, Netherlands
Expected Free Energy-based Informative Path Planning for Robotic Mars Exploration
Aug 14, 2026Ajith Anil Meera, Pablo Lanillos, Wouter KouwExpected Free Energy Informative Path Planning
TU Eindhoven, The Netherlands · Cajal Centre for Neuroscience, Spanish National Research Council, Madrid, Spain
Expected Free Energy as Belief-Dependent Utility for rho-POMDPs
Jul 18, 2026Patrick Cooper, Alvaro VelasquezFree Energy Calculation Active Inference
Department of Computer Science, University of Colorado Boulder, Boulder, CO, USA
Principled Direction-Free Intrinsic Motivation through Model-Free Epistemic Free-Energy Estimators
Jul 18, 2026Alireza Furutanpey, Schahram DustdarFree Energy Calculation Intrinsic Reward Methods
Coovally, Barcelona, Spain · Distributed Systems Group, TU Vienna, Austria · ICREA Barcelona, Spain
Free Energy Heuristics: Fast-And-Frugal Cognition as Active Inference Under Uncertain Precision
Jun 14, 2026Alex BogdanCognitive Modeling Active Inference
Evolutionairy AI Toronto, Canada
Expected Free Energy-based Planning as Variational Inference
Jun 9, 2026Wouter W. L. Nuijten, Thijs van de Laar, Bert de VriesFree Energy Calculation Active Inference
Eindhoven University of Technology, Eindhoven, the Netherlands · Lazy Dynamics B.V., the Netherlands
What Type of Inference is Active Inference?
Jun 3, 2026Wouter W. L. Nuijten, Mykola Lukashchuk, Thijs van de Laar +1Active Inference Free Energy Principle
Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands · Lazy Dynamics, Utrecht, the Netherlands
Curvature-aware Expected Free Energy as an Acquisition Function for Bayesian Optimization
Mar 27, 2026Ajith Anil Meera, Wouter KouwFree Energy Calculation Bayesian Optimization
Faculty of Electrical Engineering, TU Eindhoven, The Netherlands

Oct 20, 2026J/K move · Enter open · S save

Knowing When to Stop Thinking: A Dissection of Expected Free Energy as a Test-Time Compute Controller for Latent World-Model Planning

Tong Zhang, Zhiyuan Shi, Yun Peng, Tao Xie

Fudan University · Peking University

Expected Free Energy Inference-Time Compute Allocation World Model-Based Planning Efficient Inference for World Action Models Latent-Space Planning

Open paper AskPDF arXiv

A planner using a learned world model can improve its predicted outcome while choosing a worse action. We study when expected free energy (EFE) can help decide how much computation to spend on a plan. An EFE-inspired objective combines goal distance with disagreement among stochastic predictions. We test candidate scoring, stopping, and online predictor updates separately. Stopping when elite cost stalls reduces refinement on tasks whose success plateaus with depth, with mean success close to fixed-depth references. At the larger tested weights, rewarding disagreement lowers success with a frozen model. Penalizing it helps at some task and weight settings but harms others. When training data omit a region needed to reach the goal, online updates recover part of the resulting loss. The tested cost-stall rule can fail when further optimization exploits model errors, because predicted cost keeps falling after task success peaks. Cost progress can guide computation on a plateau, but protecting performance after a peak requires information about the consequences of the chosen actions.