Metraipolis › archive › dc_ef6cc268e801

Proving and Improving the Platonic Representation Hypothesis with JEPAs — arXiv:2610.39302

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.39302
License (as archived)
CC0-1.0
Archived
2026-10-31T06:09:15.441Z
SHA-256 of the archived text
70f8618cdcd81416437703de573a64388c80fe51d3e357d2273c89448ead09f6
Stable URL of this page
https://metraipolis.com/documento/dc_ef6cc268e801

How to cite

Proving and Improving the Platonic Representation Hypothesis with JEPAs — arXiv:2610.39302 — arXiv (en), https://arxiv.org/abs/2610.39302 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef6cc268e801 (Metraipolis archive dc_ef6cc268e801, sha256 70f8618cdcd81416437703de573a64388c80fe51d3e357d2273c89448ead09f6)
BibTeX
@misc{dcef6cc268e801,
  title = {Proving and Improving the Platonic Representation Hypothesis with JEPAs — arXiv:2610.39302},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.39302},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef6cc268e801; sha256 70f8618cdcd81416437703de573a64388c80fe51d3e357d2273c89448ead09f6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39302. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Whether models trained on different modalities converge to the same representation of the world is central to AI. The Platonic Representation Hypothesis (PRH) conjectures that they do as they scale, but without a proof it offers neither a guarantee nor a recipe. We turn the hypothesis into a theorem. First, we prove a \emph{Platonic Representation Theorem}: Joint-Embedding Predictive Architectures (JEPAs) trained separately on any number of modalities learn the same representation up to one rotation whenever their positive pairs are two observations of one scene--\emph{good views}--and their misalignment is bounded otherwise. Second, we show that next-token prediction is a JEPA with \emph{bad views}, parts of a single document, which explains why language models align with vision only through scale. Third, we validate the theory from 145k captions to the texts of 940M web images, and uncover a scaling law of alignment that predicts unseen runs within seed noise. Good views need 42 times fewer training samples than next-token prediction on the same corpus to reach the same alignment, and a 109M-parameter text JEPA aligns with DINOv2 and CLIP's image encoders better than every language model we evaluated, up to 72B parameters, and than CLIP's text encoder: next-token prediction is a remarkably inefficient route to platonic representations. As a showcase, one rotation turns separately trained encoders into an image captioner (CIDEr .838 on COCO). Even that rotation needs no pair: on the latest pair-free benchmark, our text JEPA beats every zero-shot accuracy reported with language models up to 8B parameters, e.g., 64\% against at most 56\% top-5 on ImageNet-100. Positive views, and not scale alone, are thus the lever to build platonic representations.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2610.39302 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-31T06:07:08.515Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).