Metraipolis › archive › dc_a6703516dc91

IronMind: Scaling Humanoid Dexterous Manipulation via Camera-Space Ego-Centric Pretraining — arXiv:2609.39403

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39403
License (as archived)
CC0-1.0
Archived
2026-10-07T15:12:02.715Z
SHA-256 of the archived text
c098cb7aacd8892aff7d77e8230f6d078f28d5a9dca59bf26f90b325c3fc1f8c
Stable URL of this page
https://metraipolis.com/documento/dc_a6703516dc91

How to cite

IronMind: Scaling Humanoid Dexterous Manipulation via Camera-Space Ego-Centric Pretraining — arXiv:2609.39403 — arXiv (en), https://arxiv.org/abs/2609.39403 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a6703516dc91 (Metraipolis archive dc_a6703516dc91, sha256 c098cb7aacd8892aff7d77e8230f6d078f28d5a9dca59bf26f90b325c3fc1f8c)
BibTeX
@misc{dca6703516dc91,
  title = {IronMind: Scaling Humanoid Dexterous Manipulation via Camera-Space Ego-Centric Pretraining — arXiv:2609.39403},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39403},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a6703516dc91; sha256 c098cb7aacd8892aff7d77e8230f6d078f28d5a9dca59bf26f90b325c3fc1f8c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39403. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Egocentric human video offers a scalable data source for dexterous manipulation, yet using it to train humanoid robots presents two challenges: (1) an embodiment gap, as human hands differ structurally from robot end-effectors and low-cost egocentric recordings lack the torso kinematics required by conventional retargeting; and (2) heterogeneous data quality, including noisy hand-pose tracking and weakly aligned text annotations. We introduce IronMind, a vision-language-action (VLA) model that uses egocentric human video and heterogeneous robot data to pretrain policies for humanoid dexterous manipulation. To bridge the embodiment gap, IronMind bypasses explicit body-retargeting by using a camera-space action representation, the native reference space of egocentric video, and semantically aligning robot and human action dimensions. Across total pretraining budgets from 250 to 10,000 hours, validation loss decreases approximately log-linearly with data scale. Larger pretraining budgets also improve out-of-distribution real-robot manipulation after post-training: across six challenging tasks with unseen objects, affordances, and reasoning prompts, the 10,000-hour model achieves a 55.0% success rate, compared with at most 11.7% for every pretraining budget up to 5,000 hours and 5.0% without pretraining. At the same pretraining budget, the camera-space action representation also outperforms the torso-frame baseline. Together, these findings support pretraining with a camera-space action representation on large-scale human egocentric data as a scalable foundation for humanoid robot manipulation.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-01 in cs.RO).
Categorie arXiv: Robotics (cs.RO).
Pagina del lavoro: https://arxiv.org/abs/2609.39403 — 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 2026-10-07T14:44:47.441Z dalla pagina https://arxiv.org/catchup/cs.RO/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).