Publications

Journal articles

2026 · Scientific Reports

An elementary form of agency at planetary-scale

Sergio Rubin, Conor Heins, Takahito Mitsui, Lancelot Da Costa, Karl Friston

A Bayesian mechanics description of coupled biosphere–climate dynamics as planetary-scale active inference.

link doi
2025 · Frontiers in Network Physiology, 5:1521963

From Pixels to Planning: Scale-Free Active Inference

Karl Friston, Conor Heins, Tim Verbelen, Lancelot Da Costa, Tommaso Salvatori, Dimitrije Markovic, Alexander Tschantz, Magnus Koudahl, Christopher Buckley, Thomas Parr

link doi
2024 · PNAS, 121(17): e2320239121

Collective Behavior from Surprise Minimization

Conor Heins, Beren Millidge, Lancelot Da Costa, Richard Mann, Karl Friston, Iain Couzin

The first active inference model of collective motion — surprise-minimizing agents spontaneously reproduce real-world flocking patterns.

link doi
2022 · Journal of Open Source Software, 7(73): 4098

pymdp: A Python Library for Active Inference in Discrete State-Spaces

Conor Heins, Beren Millidge, Dimitris Demekas, Brennan Klein, Karl Friston, Iain Couzin, Adam Tschantz

link doi code
2020 · Frontiers in Artificial Intelligence, 3:81

Deep Active Inference and Scene Construction

Conor Heins, Mehdi Mirza, Thomas Parr, Karl Friston, Igor Kagan, Arezoo Pooresmaeili

link doi

Preprints

2026 · arXiv preprint

Variational Filtering for Bayesian Last Layers with an Independent Binary Probit Surrogate

Dimitrije Markovic, Conor Heins, Alexander Tschantz, Christopher L. Buckley

Online variational inference for Bayesian last layers using an independent binary probit surrogate.

arXiv pdf code
2025 · arXiv preprint

AXIOM: Learning to Play Games in Minutes with Expanding Object-Centric Models

Conor Heins, Toon Van de Maele, Alexander Tschantz, Hampus Linander, Dimitrije Markovic, Tommaso Salvatori, Corrado Pezzato, Ozan Catal, Ran Wei, Magnus Koudahl, et al.

Object-centric model-based agent that learns Atari-like games in minutes via online structure learning + MPPI planning.

link arXiv
2025 · arXiv preprint

Soft Geometric Inductive Bias for Object-Centric Dynamics

Hampus Linander, Conor Heins, Alexander Tschantz, Marco Perin, Christopher Buckley

Geometric transformer world models for 2-D / 3-D trajectories and the Waymo Open Motion Dataset.

link arXiv code
2024 · arXiv preprint

Gradient-Free Variational Learning with Conditional Mixture Networks

Conor Heins, Han Wu, Dimitrije Markovic, Alexander Tschantz, Jeff Beck, Christopher Buckley

link arXiv
2024 · arXiv preprint

Navigation Under Uncertainty: Trajectory Prediction and Occlusion Reasoning with Switching Dynamical Systems

Rui Wei, Jason Lee, Sho Wakayama, Alexander Tschantz, Conor Heins, Christopher Buckley, John Carenbauer, Hari Thiruvengada, Manuel Albarracin, Miguel de Prado, Peter Horling, Patrik Winzell, Ram Rajagopal

link arXiv

Workshop papers

2026 · 7th International Workshop on Active Inference (IWAI 2026), accepted

Collective Sensing as Emergent Bayesian Inference

Conor Heins

A Bayesian description of collective sensing, in which relationships between fish positions and swimming speeds encode information about environmental gradients.

arXiv pdf interactive essay & simulation