Metraipolis › archive › dc_a03003ebafdb

Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection — arXiv:2610.27364

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.27364
License (as archived): CC0-1.0
Archived: 2026-10-21T07:07:42.118Z
SHA-256 of the archived text: d2e5720c59b758825d0f417b136d68abb72d232daf54fa15dec9e0b49c9adbac
Stable URL of this page: https://metraipolis.com/documento/dc_a03003ebafdb

How to cite

Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection — arXiv:2610.27364 — arXiv (en), https://arxiv.org/abs/2610.27364 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a03003ebafdb (Metraipolis archive dc_a03003ebafdb, sha256 d2e5720c59b758825d0f417b136d68abb72d232daf54fa15dec9e0b49c9adbac)

BibTeX

@misc{dca03003ebafdb,
  title = {Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection — arXiv:2610.27364},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.27364},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a03003ebafdb; sha256 d2e5720c59b758825d0f417b136d68abb72d232daf54fa15dec9e0b49c9adbac},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27364. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In joint-embedding predictive architectures (JEPA), an encoder and a predictor are trained together to produce a representation that minimizes the error in predicting the representation of the next observation. Because the representations of both the current and the next observation are learned, it is unclear which features they will include. Earlier characterizations describe the features JEPA selects in terms of their individual predictability, raising the concern that unpredictable features, like a traffic light, will not be encoded into the representation. However, the prediction error is also affected by the relationships between features, e.g., the light's color helps predict whether the traffic will stop. We propose the predictive closure characterization, which accounts for these relationships. Based on a reformulation of the JEPA training objective, we show that selection depends on properties of feature sets. This view agrees with earlier works when features evolve independently, but makes new predictions when they do not, for instance that a feature like the traffic light can still be retained. We test this characterization in controlled experiments, which show that unpredictable features can become better represented when they help predict other features, and that changing one feature's predictive relationships can affect how well others are represented. We also demonstrate implications for steering feature selection: targeting one feature with an auxiliary objective, or supplying it to the predictor as side information, can strengthen the representation of another feature whose future values it helps predict, so steering can be indirect.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.27364 — 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 2026-10-21T07:05:43.554Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).