Metraipolis › archive › dc_bb9ef565d85f

CAPABLE: Capability-Aware Policy Adaptation via Behavioral Latent Encoding — arXiv:2610.11971

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.11971
License (as archived): CC0-1.0
Archived: 2026-10-09T06:24:16.815Z
SHA-256 of the archived text: 5d9f7cf79bd6b7dd6efca27dd2cd990ccdfeb254180f96ffaabe4b4cf079b425
Stable URL of this page: https://metraipolis.com/documento/dc_bb9ef565d85f

How to cite

CAPABLE: Capability-Aware Policy Adaptation via Behavioral Latent Encoding — arXiv:2610.11971 — arXiv (en), https://arxiv.org/abs/2610.11971 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bb9ef565d85f (Metraipolis archive dc_bb9ef565d85f, sha256 5d9f7cf79bd6b7dd6efca27dd2cd990ccdfeb254180f96ffaabe4b4cf079b425)

BibTeX

@misc{dcbb9ef565d85f,
  title = {CAPABLE: Capability-Aware Policy Adaptation via Behavioral Latent Encoding — arXiv:2610.11971},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.11971},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bb9ef565d85f; sha256 5d9f7cf79bd6b7dd6efca27dd2cd990ccdfeb254180f96ffaabe4b4cf079b425},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11971. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-language-action (VLA) policies assume the embodiment on which they were trained and can fail when a joint fault changes how commanded actions are physically executed. Existing fault-recovery methods often require task-specific retraining, fault labels, explicit diagnosis, or privileged embodiment information. We introduce CAPABLE, a unified capability-aware adaptation framework for frozen VLAs that integrates self-supervised capability inference with residual reinforcement learning. CAPABLE infers capability, how much of the commanded motion each joint actually realizes and how that motion contributes to end-effector behavior, online from command-response history and kinematics using a temporal encoder shared across joints, Jacobian grounding, cross-joint attention, and self-supervised physical prediction. The resulting representation conditions a residual policy that adds bounded corrections to the VLA arm action without fault labels or faulty-joint identifiers. Across 28 LIBERO tasks, CAPABLE raises success on an actuator excluded from fault training from 24.8% to 59.3%, outperforming a parameter-matched global-history baseline by 17.4 points while preserving healthy performance. Leave-one-actuator-out experiments across six joints show that this transfer is not specific to one actuator, and additional evaluations characterize transfer to unseen fault families and demonstrate recovery on a physical Franka Panda. this https URL

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO) ; Machine Learning (cs.LG); Systems and Control (eess.SY).
Pagina del lavoro: https://arxiv.org/abs/2610.11971 — 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 2026-10-09T05:50:32.200Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).