JEPA · 2026-11-13

Same-State Pairs Make One JEPA Proof of the Platonic Representation Hypothesis Vacuous

Xin Li

arXiv:2611.16182PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

The authors show that, under the hard-whitening JEPA population objective with same-state positive pairs, the induced world kernel becomes the identity, so every centered whitened state function is a global minimizer and rotation-uniqueness becomes vacuous. They further show an orbit-level failure: two minimizers can reach the maximum best-rotation residual. One surviving result is a gap-free lower bound on alignment error, but it cannot certify good-view alignment in the degenerate regime.

Why it matters

The paper corrects a conditional JEPA proof of the Platonic Representation Hypothesis by arguing that its uniqueness premises do not hold in the same-state regime relevant to the motivating evidence. This means an “agreement up to rotation” number can reflect shared subspace alignment with an unresolved rotation gauge rather than the uniqueness claim. The authors also provide a resource–gauge ledger describing which pairing, noise, anchors, and constraints fix continuous vs discrete gauge freedoms.

Method

  • Hard-whitening population objective with same-state positive pairs is analyzed via the induced world kernel Kπ and Markov operator Tπ on centered functions.
  • Orbit-level and eigenstructure arguments construct global minimizers whose best-rotation squared residual attains the maximum possible under dimension conditions.
  • They isolate the surviving alignment-error lower bound (gap-free) and explain identifiability needs: fixing a continuous O(k) gauge and a discrete signed-permutation gauge via anchors/pairing resources.

Limitation

The authors state that the gap-assuming uniqueness theorem is conditional: for 1 ≤ k < d_z − 1 the uniqueness premise fails under the same-state regime, and the paper’s motivating experiments leave the injective-rendering premise unestablished.

Abstract (from arXiv)

In a Joint-Embedding Predictive Architecture (JEPA), the JEPA proof of the Platonic Representation Hypothesis (PRH) assumes that every observation determines the world state and that a spectral gap makes minimizers unique up to rotation. Under the hard-whitening population objective, same-state positive pairs make the induced world kernel the identity: every centered whitened state function is a global minimizer, and, when $d_z-1\ge 2k$, two minimizers can have best-rotation squared residual $2k$. The uniqueness claim is conditional; for $k<d_z-1$ its idealized same-state limit is vacuous. One of the paper's four quantitative results survives: a gap-free lower bound. Its bound vanishes only at perfect scene-explained fraction; it bounds alignment error but cannot certify good-view alignment. A canonical frame is determined only up to a rotation and a latent permutation. Paired samples fix the relative frame at effective rank $k$; a global rotation remains. Component semantics fix the relative permutation only up to signs; value anchors fix it once the mixing frame is pinned. A resource--gauge ledger records which resource fixes which gauge.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-11-12 · Indexed by Pub-AI: 2026-11-13 UTC

preprint

Same-State Pairs Make One JEPA Proof of the Platonic Representation Hypothesis Vacuous

Xin Li

Preprint posted: 2026-11-12 · Original record ↗

arxiv
2611.16182
Explicit source versions (1)
  • v1 · 2026-11-12
  1. Preprint available

    Source date: 2026-11-12 · First detected: 2026-11-13 UTC · Historical baseline

    Source evidence (2)
Abstract · arxiv

In a Joint-Embedding Predictive Architecture (JEPA), the JEPA proof of the Platonic Representation Hypothesis (PRH) assumes that every observation determines the world state and that a spectral gap makes minimizers unique up to rotation. Under the hard-whitening population objective, same-state positive pairs make the induced world kernel the identity: every centered whitened state function is a global minimizer, and, when $d_z-1\ge 2k$, two minimizers can have best-rotation squared residual $2k$. The uniqueness claim is conditional; for $k<d_z-1$, where $d_z$ is the number of world states, this limit makes its premise unsatisfiable, so the claim has no content there. The experiments that motivate it leave the assumption: a caption sees only part of an image, and the paper states no experiment meets its hypotheses exactly. The theorem does not reach those experiments, and for $k<d_z-1$ its idealized same-state limit is vacuous. One of the paper's four quantitative results survives: a gap-free lower bound. Its bound vanishes only at perfect scene-explained fraction; it bounds alignment error but cannot certify good-view alignment. A canonical frame is determined only up to a rotation and a latent permutation. Paired samples fix the relative frame at effective rank $k$; a global rotation remains. Component semantics fix the relative permutation only up to signs; value anchors fix it once the mixing frame is pinned. A resource--gauge ledger records which resource fixes which gauge.