Emergent Geometry in Neural Representations of the Visual World

Liam Storan, Nikos Karantzas, Klindt, D., Willeke, K. F., Deng, D., Michał Gerasimiuk, Xaq Pitkow, Franke, K., Miolane, N., Sanborn, S., & Tolias, A. S.

Using a digital twin of primate V4, we show that neural representations in vision mirror the geometric structure of the visual world, enabling equivariant encodings that disentangle latent factors like identity and orientation and support strong generalization across natural contexts—paralleling principles observed in artificial neural networks.