Metraipolis › archive › dc_dcf5e9a758de

ARC-Bench: Closed-Loop Replanning Masks Broken Action Ranking in Frozen JEPA World Models — arXiv:2609.05461

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.05461
License (as archived)
CC0-1.0
Archived
2026-10-16T13:58:34.899Z
SHA-256 of the archived text
1bebd0857cce3c527d59d1a6afe294f849448b590d6be2197fa688f2e8d06146
Stable URL of this page
https://metraipolis.com/documento/dc_dcf5e9a758de

How to cite

ARC-Bench: Closed-Loop Replanning Masks Broken Action Ranking in Frozen JEPA World Models — arXiv:2609.05461 — arXiv (en), https://arxiv.org/abs/2609.05461 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dcf5e9a758de (Metraipolis archive dc_dcf5e9a758de, sha256 1bebd0857cce3c527d59d1a6afe294f849448b590d6be2197fa688f2e8d06146)
BibTeX
@misc{dcdcf5e9a758de,
  title = {ARC-Bench: Closed-Loop Replanning Masks Broken Action Ranking in Frozen JEPA World Models — arXiv:2609.05461},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.05461},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dcf5e9a758de; sha256 1bebd0857cce3c527d59d1a6afe294f849448b590d6be2197fa688f2e8d06146},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.05461. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reward-free latent world models plan by scoring candidate actions with distances in a frozen latent space: an action is preferred if its predicted future embedding lands closer to the goal embedding. This silently assumes that latent closeness is action-rankable, i.e., that ordering candidates by latent distance agrees with ordering them by true cost. We audit this assumption directly. We introduce ARC-Bench, a no-leak, fixed-candidate protocol that measures whether frozen JEPA-style objectives rank candidate actions correctly, and apply it to official released JEPA-WM checkpoints across navigation and manipulation-style control. The assumption fails, severely and structurally: on the official manipulation audits the top-scored candidate is almost always suboptimal, and the same inversion appears in the maze domains. A controlled visual-backbone extension shows that the defect persists when DINOv2 is replaced by video-pretrained V-JEPA 1 and V-JEPA 2 encoders at ViT-L/ViT-G scale. Provenance, undertraining, matched-budget backbone controls, and metric-circularity controls rule out trivial explanations. We then explain why this defect has stayed invisible: closed-loop replanning masks it. When we reduce the planner's replanning frequency, success collapses in both a navigation and a manipulation domain, and the episodes rescued by frequent replanning are enriched for severe first-plan ranking failures in the PointMaze first-plan diagnostic. Closed-loop success rates therefore systematically overstate the rankability of frozen latent representations. ARC-Bench supplies the measurement, and the masking mechanism the explanation, for methods that adapt, amortize, or replan around latent-space planners without directly auditing released JEPA-WM action rankability.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.LG).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG); Robotics (cs.RO).
Pagina del lavoro: https://arxiv.org/abs/2609.05461 — 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 2026-10-16T13:39:43.106Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).