Manu Halvagal

We think cortex might function like a JEPA. It looks like prediction errors in layer 2/3 are not computed against input (as is the idea in predictive processing), but against a representation in latent space (i.e. like in a JEPA arxiv.org/abs/2301.08243 or RPL doi.org/10.1101/2025...).

Thanks Rui and Friedemann. The model in Nejad et. al. (2025) also contained a feedforward connection between L2/3 and L5, which muddles the mapping to a JEPA a fair bit imo

glad to share our new preprint on how representation learning in the cortex can be explained by prediction across time in a recurrent JEPA. check it out here 👇