The Platonic Universe: Do Foundation Models See the Same Sky?

2025-09-23

AI Summary

This study empirically evaluates the Platonic Representation Hypothesis (PRH) in astronomy—whether diverse deep learning architectures converge to a shared, physics-informed latent representation of galaxies. Method: We conduct cross-model mutual k-nearest neighbor (mKNN) analysis across vision Transformers, self-supervised models, and astronomy-specific architectures, trained on multi-source observational data including JWST. Contribution/Results: We find that larger model capacity strongly correlates with higher representational alignment, supporting convergence toward a unified astrophysical representation space. Crucially, general-purpose pre-trained architectures significantly outperform domain-specific designs, validating their suitability as astronomical foundation models. This work provides the first galaxy-scale empirical confirmation of PRH, establishing both theoretical grounding and a methodological framework for leveraging large-scale AI infrastructure to build generalizable, physics-aware astronomical intelligence.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Representation Learning for Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

We test the Platonic Representation Hypothesis (PRH) in astronomy by measuring representational convergence across a range of foundation models trained on different data types. Using spectroscopic and imaging observations from JWST, HSC, Legacy Survey, and DESI, we compare representations from vision transformers, self-supervised models, and astronomy-specific architectures via mutual $k$-nearest neighbour analysis. We observe consistent scaling: representational alignment generally increases with model capacity across our tested architectures, supporting convergence toward a shared representation of galaxy astrophysics. Our results suggest that astronomical foundation models can use pre-trained general-purpose architectures, allowing us to capitalise on the broader machine learning community's already-spent computational investment.

Problem

• Testing the Platonic Representation Hypothesis in astronomy using foundation models
• Measuring representational convergence across models trained on different astronomical data
• Determining if foundation models develop shared galaxy astrophysics representations

Innovation

• Testing representational convergence across foundation models
• Comparing vision transformers and self-supervised architectures
• Observing alignment scaling with model capacity