Platonic Representations in the Human Brain: Unsupervised Recovery of Universal Geometry

2026-05-19

AI Summary

This study investigates whether a shared neural representational geometry exists across human brains and demonstrates how to align brain activity representations across subjects without paired data. Leveraging fMRI recordings acquired during naturalistic viewing, the work employs a self-supervised encoder to learn subject-specific brain representations and then maps them into a common latent space via an unsupervised orthogonal transformation. This approach is the first to recover—without supervision, paired samples, or intermediary models—an approximately isometric, universal representational geometry in the human brain. Experimental results show that the proposed framework substantially improves cross-subject retrieval accuracy of brain representations, providing strong evidence that the human visual cortex harbors a shared neural geometric structure compatible with a unified coordinate system.

Abstract

The Strong Platonic Representation Hypothesis suggests that representational convergence in artificial neural networks can be harnessed constructively: embeddings can be translated across models through a universal latent space without paired data. We ask whether an analogous geometry can be recovered across human brains. Using fMRI data from the Natural Scenes Dataset, we propose a self-supervised encoder that learns subject-specific embeddings from brain data alone by exploiting repeated stimulus presentations. We show that these independently learned spaces can be translated across subjects using unsupervised orthogonal rotations, without paired cross-subject samples or intermediate model representations. Synchronizing pairwise rotations into a single shared latent space further improves cross-subject retrieval, indicating that subject-specific spaces are mutually compatible with a common coordinate system. These results provide evidence for a shared neural geometry in the human visual cortex: subject-specific fMRI representations are approximately isometric across individuals and can be translated through purely geometric transformations.

Problem

• Platonic representation
• neural geometry
• cross-subject alignment
• fMRI
• representational convergence

Innovation

• Platonic representations
• unsupervised alignment
• neural geometry
• fMRI embeddings
• orthogonal rotation