Gestalt: a meta-foundation model for astronomy
Abstract
The Platonic Representation Hypothesis predicts that sufficiently scaled foundation models converge on a shared representation of the world. As each non-converged model gives a noisy view of a common structure when passed the same input, we ask whether we can combine models into a representation that outperforms its individual components. We test this on galaxies: we embed images via a basket of 22 frozen foundation models from eight families, whiten each view, and take a randomised SVD of the embedding concatenation. The resulting 1024-dimensional embedding outperforms every basket member on 19/21 of our tested metrics for physical property and galaxy morphology estimation for HSC, JWST, and DESI Legacy Survey imagery. We find that performance rises with basket size and basket architectural diversity, and that the meta-foundation model's performance transfers across astronomical surveys. We conclude that a useful astronomical foundation model can be assembled from existing generalist models with no training required beyond a single unsupervised projection. By leveraging the community's already-spent work, we save a lot of compute: a fresh pre-train of a comparable single-domain model would cost 𝒪(104--105) A100 GPU hours (emitting several tonnes of CO2eq.), whereas assembling Gestalt requires minutes on a single machine.
Related
- DINOspec: Efficient Multimodal Alignment of Vision and Spectral Foundation Models for Astronomy
- Transitioning from ADS to SciX to Serve 21st Century Astronomy
- From Axion Detection to Dark Matter Astronomy
- Neutrino Astronomy at High Energies - An Experimental Review
- The Promise of Gravitational-Wave Astronomy
BibTeX
@misc{smith2026gestalt,
title = {{Gestalt: a meta-foundation model for astronomy}},
author = {Michael J. Smith and Shashwat Sourav},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.38312},
eprint = {2609.38312},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.38312},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Gestalt: a meta-foundation model for astronomy
AU - Michael J. Smith
AU - Shashwat Sourav
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.38312
UR - https://arxiv.org/abs/2609.38312
AB - The Platonic Representation Hypothesis predicts that sufficiently scaled foundation models converge on a shared representation of the world. As each non-converged model gives a noisy view of a common structure when passed the same input, we ask whether we can combine models into a representation that outperforms its individual components. We test this on galaxies: we embed images via a basket of 22 frozen foundation models from eight families, whiten each view, and take a randomised SVD of the embedding concatenation. The resulting 1024-dimensional embedding outperforms every basket member on 19/21 of our tested metrics for physical property and galaxy morphology estimation for HSC, JWST, and DESI Legacy Survey imagery. We find that performance rises with basket size and basket architectural diversity, and that the meta-foundation model's performance transfers across astronomical surveys. We conclude that a useful astronomical foundation model can be assembled from existing generalist models with no training required beyond a single unsupervised projection. By leveraging the community's already-spent work, we save a lot of compute: a fresh pre-train of a comparable single-domain model would cost 𝒪(10^4--10^5) A100 GPU hours (emitting several tonnes of CO_2eq.), whereas assembling Gestalt requires minutes on a single machine.
ER -