# The Platonic Universe: Do Foundation Models See the Same Sky? arXiv: https://arxiv.org/abs/2509.19453 · HTML: https://mlsanity.com/paper/2509.19453/ ## Key facts - Cited by 6 papers in the map (data through 14 Oct 2026). - Builds on 31 of its 76 references that are in the map. - First posted on arXiv 23 Sep 2025. ## Builds on (most cited first, 30 of 31) - [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021) - [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020) - [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014) - [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023) - [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023) - [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021) - [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021) - [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020) - [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023) - [Training Compute-Optimal Large Language Models](https://mlsanity.com/paper/2203.15556/) (2022) - [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019) - [DINOv3](https://mlsanity.com/paper/2508.10104/) (2025) - [Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture](https://mlsanity.com/paper/2301.08243/) (2023) - [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025) - [Large Language Models](https://mlsanity.com/paper/2307.05782/) (2023) - [ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders](https://mlsanity.com/paper/2301.00808/) (2023) - [PaliGemma: A versatile 3B VLM for transfer](https://mlsanity.com/paper/2407.07726/) (2024) - [Understanding image representations by measuring their equivariance and equivalence](https://mlsanity.com/paper/1411.5908/) (2014) - [Convergent Learning: Do different neural networks learn the same representations?](https://mlsanity.com/paper/1511.07543/) (2015) - [Revisiting Model Stitching to Compare Neural Representations](https://mlsanity.com/paper/2106.07682/) (2021) - [Rotation-invariant convolutional neural networks for galaxy morphology prediction](https://mlsanity.com/paper/1503.07077/) (2015) - [AstroLLaMA: Towards Specialized Foundation Models in Astronomy](https://mlsanity.com/paper/2309.06126/) (2023) - [Realistic galaxy image simulation via score-based generative models](https://mlsanity.com/paper/2111.01713/) (2021) - [Astronomia ex machina: a history, primer, and outlook on neural networks in astronomy](https://mlsanity.com/paper/2211.03796/) (2022) - [PAPERCLIP: Associating Astronomical Observations and Natural Language with Multi-Modal Models](https://mlsanity.com/paper/2403.08851/) (2024) - [AstroLLaMA-Chat: Scaling AstroLLaMA with Conversational and Diverse Datasets](https://mlsanity.com/paper/2401.01916/) (2024) - [AstroPT: Scaling Large Observation Models for Astronomy](https://mlsanity.com/paper/2405.14930/) (2024) - [AstroLLaVA: towards the unification of astronomical data and natural language](https://mlsanity.com/paper/2504.08583/) (2025) - [Predicting galaxy spectra from images with hybrid convolutional neural networks](https://mlsanity.com/paper/2009.12318/) (2020) - [SpecCLIP: Aligning and Translating Spectroscopic Measurements for Stars](https://mlsanity.com/paper/2507.01939/) (2025) ## Built on by (most cited first, 6 of 6) - [ENIGMA: The Geometry of Reasoning and Alignment in Large-Language Models](https://mlsanity.com/paper/2510.11278/) (2025) - [Scalable and Interpretable Representation Alignment with Ordinal Similarity](https://mlsanity.com/paper/2606.16379/) (2026) - [Do Foundation Models Know Geometry? Probing Frozen Features for Continuous Physical Measurement](https://mlsanity.com/paper/2603.06459/) (2026) - [What AstroPT knows about galaxies, and what that can teach us about LLMs](https://mlsanity.com/paper/2608.22614/) (2026) - [Gestalt: a meta-foundation model for astronomy](https://mlsanity.com/paper/2609.38312/) (2026) - [Shared Geometry Is Not Shared Physics: A Layerwise Test of the Platonic Representation Hypothesis in Astronomy](https://mlsanity.com/paper/2610.04130/) (2026) ## Abstract (arXiv) We investigate when foundation models converge towards shared representations, and how this convergence depends on model capacity, training regime, and model architecture. We take a `science-for-AI' approach, using astronomy as an experimental instrument to test the Platonic Representation Hypothesis and its Aristotelian refinement against an external physical reference. The historical success of astrophysics is evidence that a compact, modality-invariant description of galaxy observables exists, and so representation convergence toward reality should be measurable against the physical parameters astronomers already use. Given this framework, we evaluate eleven foundation model families (spanning classification, self-distillation, joint-embedding prediction, autoencoding, vision-language pre-training, and astro-specific architectures from O(10M)→O(10B) parameters) on crossmatched JWST, HSC, and Legacy imagery, and DESI spectroscopy. All models are evaluated frozen, with no astronomy-specific fine-tuning. We probe redshift, stellar mass, and sSFR via linear probes, and local (MKNN) and global (CKA) embedding geometry within families, between modalities, and across architectures. We find that physics performance scales predictably with capacity; probe directions align consistently with expected astrophysical correlations and selection effects; and local (not global) embedding alignment tracks physics performance, including between DESI spectra and HSC imagery---modalities that share essentially no low-level statistics. Our results support the ARH over the strict PRH, demonstrate astronomy's value as an experimental framework for neural representation learning, and suggest that astro-foundation models can build on general-purpose pre-trained architectures, capitalizing on the broader open machine learning community's already-spent computational investment.