Metraipolis › archive › dc_fe7d4e569204

Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency — arXiv:2608.12939

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.12939
License (as archived)
CC0-1.0
Archived
2026-10-19T04:27:30.333Z
SHA-256 of the archived text
d801c1a889d0da8a50f92858bae3c6b6f5fdeaec4eb971f6c97cb53fda3dbebb
Stable URL of this page
https://metraipolis.com/documento/dc_fe7d4e569204

How to cite

Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency — arXiv:2608.12939 — arXiv (en), https://arxiv.org/abs/2608.12939 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fe7d4e569204 (Metraipolis archive dc_fe7d4e569204, sha256 d801c1a889d0da8a50f92858bae3c6b6f5fdeaec4eb971f6c97cb53fda3dbebb)
BibTeX
@misc{dcfe7d4e569204,
  title = {Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency — arXiv:2608.12939},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.12939},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fe7d4e569204; sha256 d801c1a889d0da8a50f92858bae3c6b6f5fdeaec4eb971f6c97cb53fda3dbebb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.12939. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Joint-embedding predictive architectures (JEPAs) learn world models that predict in a compact latent space rather than in pixels, reducing the pressure to model nuisance appearance. Yet this provides no guarantee against visual perturbations: they can still alter the encoded representation and affect subsequent action-conditioned predictions. Bisimulation captures this requirement precisely: two observations should be treated as the same state only when their action-conditioned consequences agree. Guided by this criterion, we introduce Action-Conditioned Predictive Consistency (ACPC), a diagnostic that measures how far a clean history and a visually perturbed view of it diverge after being rolled forward under the same action sequence. We prove that this divergence bounds the perturbation-induced change in multi-step prediction error and planner cost. Building on pairwise ACPC, we define two complementary measures: the Invariance Radius (IR) summarizes clean-perturbed rollout spread, while the Separation Rate (SR) checks whether different states remain distinguishable after rollout. Experiments on four visual control tasks show that pairwise ACPC predicts perturbation-induced prediction and cost changes. On LeWM, the IR-SR screen transfers across tasks, and the joint diagnostic remains informative under blur and resize. PLDM exhibits similar diagnostic trends under a different architecture.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-14 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.12939 — 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 2026-10-19T04:20:35.836Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-14?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).