Metraipolis › archive › dc_327f1a352bb3

FastOPD: On-Policy Distillation for Lightweight VLA Deployment — arXiv:2610.02832

Source
arXiv
Original
https://arxiv.org/abs/2610.02832
License (as archived)
CC0-1.0
Archived
2026-10-05T22:50:17.419Z
SHA-256 of the archived text
7dccd0b49b349f1647d3a9594e4cfed3a765fe10eb021e76df258bd896ed7431
Stable URL of this page
https://metraipolis.com/documento/dc_327f1a352bb3

How to cite

FastOPD: On-Policy Distillation for Lightweight VLA Deployment — arXiv:2610.02832 — arXiv, https://arxiv.org/abs/2610.02832 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_327f1a352bb3 (Metraipolis archive dc_327f1a352bb3, sha256 7dccd0b49b349f1647d3a9594e4cfed3a765fe10eb021e76df258bd896ed7431)
BibTeX
@misc{dc327f1a352bb3,
  title = {FastOPD: On-Policy Distillation for Lightweight VLA Deployment — arXiv:2610.02832},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.02832},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_327f1a352bb3; sha256 7dccd0b49b349f1647d3a9594e4cfed3a765fe10eb021e76df258bd896ed7431},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02832. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-Language-Action (VLA) foundation models have scaled rapidly to enhance manipulation performance and generalizability, but this scaling incurs high computational costs that render real-world deployment increasingly challenging. Existing approaches typically mitigate this issue by designing smaller architectures or reducing the iterative denoising steps in flow-based policies. In this work, we propose FastOPD, a foundation-to-lightweight VLA framework that enables the practical deployment of large-scale VLAs through efficient on-policy distillation. Specifically, FastOPD adapts a flow map for single-state teacher supervision and combines it with a self-consistency objective to construct a compact student that learns the teacher dynamics. Furthermore, we theoretically demonstrate that minimizing this objective allows the distilled student to recover a distribution on par with that induced by an ideal few-step teacher model. We evaluate FastOPD across diverse foundation policies in simulation and real-world experiments. On LIBERO, FastOPD retains 84% of the performance of $\pi_{0.5}$ with only two inference steps, reducing inference latency by 78.1% while outperforming existing few-step distillation baselines in average success rate. With LingBot-VLA as the teacher, FastOPD improves the single-step success rate over the base student by 15.9 percentage points on RoboTwin 2.0. We further demonstrate its applicability to a World Action Model (WAM) and deploy a compact student distilled from MolmoAct2 on a real robot.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.02832 — 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 2026-10-05T22:45:55.800Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).