Our paper with @eplantec.bsky.social was accepted at @alifeconf.bsky.social
How does criticality move from agents to populations?
We study how near-critical dynamics inside individual agents shape collective dynamics at the population scale.
arxiv.org/pdf/2605.01818
New research out — A developmental model with morphogenetic scaffoldings to guide self-organisation: a single model, many grown patterns. The memory-compute trade-off but in self-organizing systems.
A collab featuring @najarro.science, Jakob Schauser and @risi.bsky.social and yours truly
We are thrilled to be returning to GECCO for a second edition of the Evolving Self-Organisation Workshop.
We are now accepting paper submissions, so come join us!
Check out our website for more info and upcoming announcements evolving-self-organisation-workshop.github.io/gecco-2026/
Video
Video
Video
7/n Thus we show that the system can offer a valuable insight related to the functional utility of pre-patterns. Furthermore we can use it to replicate how embryogenesis combines growth and self-organisation.
Check out the preprint here: arxiv.org/abs/2605.14998
Milton
nbes
Milton
Milton
5/n Using a goal-directed NCA formulation, we show that for similar parameter counts, using pre-patterns leads to systems that can encode more patterns:
3/n Our results show 3 things:
1. Pre-patterns improve robustness,
2. they increase model capacity,
3. and, perhaps surprisingly, are not used to directly approximate the target (or some approximation of it)
2/n To study this question we paired an NCA with a learned spatial pre-pattern generator in the form of an implicit representation model, specifically a SIREN.
The pre-pattern acts as a scaffolding which sets initial conditions of the NCA but cannot reproduce the target directly
1/n Self-organisation is studied as a mechanism that works in isolation, unfolding development from random patterns. But biology rarely works this way; embryos start with spatial information guiding development. In this work we model this computationally arxiv.org/abs/2605.14998
6/n Most surprisingly, we find that the pre-patterns are not used to directly approximate the target.
Indeed, even though more complex SIREN functions support better target generation, the initial patterns are not closer to the target when we increase complexity
4/n To test the first result we trained the model using different levels of reliability in the cell update and compared to a standard NCA which can only self-organise. The results show that as we decrease cell reliability, pre-patterns help with robustness:
