Metraipolis › archive › dc_69d54d6528fa

LEAP: Making Privileged Geometry Supervision Effective for Visuomotor Learning — arXiv:2610.07015

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.07015
License (as archived): CC0-1.0
Archived: 2026-10-07T13:11:42.571Z
SHA-256 of the archived text: 38b078991adbc56f39838ce3a0f4b1d485980f22cdf6dbaabbfd10083c4c3adf
Stable URL of this page: https://metraipolis.com/documento/dc_69d54d6528fa

How to cite

LEAP: Making Privileged Geometry Supervision Effective for Visuomotor Learning — arXiv:2610.07015 — arXiv (en), https://arxiv.org/abs/2610.07015 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_69d54d6528fa (Metraipolis archive dc_69d54d6528fa, sha256 38b078991adbc56f39838ce3a0f4b1d485980f22cdf6dbaabbfd10083c4c3adf)

BibTeX

@misc{dc69d54d6528fa,
  title = {LEAP: Making Privileged Geometry Supervision Effective for Visuomotor Learning — arXiv:2610.07015},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.07015},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_69d54d6528fa; sha256 38b078991adbc56f39838ce3a0f4b1d485980f22cdf6dbaabbfd10083c4c3adf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07015. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Privileged 3D supervision uses additional geometric information during training to guide RGB-based visuomotor policy learning, without requiring geometric inputs at deployment. However, low reconstruction error does not ensure that visual representations capture geometry useful for control. We identify three limitations that weaken this supervision: proprioceptive shortcuts, dominant-view reliance, and reconstruction objectives dominated by task-irrelevant geometry. To address these limitations, we propose Latent Encoding with Aligned Privileged Geometry (LEAP). Our framework uses an auxiliary decoder to reconstruct point clouds within the manipulation workspace from visual features alone, while retaining proprioception for action prediction. Alongside full reconstruction, we introduce wrist-view dropout and partial reconstruction targets matched to the retained wrist, encouraging the encoder to capture complementary local geometry. The auxiliary decoder is removed at inference, leaving only RGB observations and robot state as policy inputs. Extensive experiments on RoboTwin, ManiSkill, and real-world tasks demonstrate consistent and substantial improvements over Diffusion Policy and ACT, with only a small increase in parameter count and no reduction in inference speed.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.RO, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO).
Pagina del lavoro: https://arxiv.org/abs/2610.07015 — 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 2026-10-07T12:42:33.496Z dalla pagina https://arxiv.org/list/cs.RO/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).