Metraipolis › archive › dc_c79be4401cf2

Fast-VLA: You Only Think Once for Robot Manipulation — arXiv:2610.21389

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21389
License (as archived)
CC0-1.0
Archived
2026-10-16T10:39:54.808Z
SHA-256 of the archived text
5409f9c168d8d2b548f836710aaf6c5af2e73954f9631ed44c75915f1f8cc6cb
Stable URL of this page
https://metraipolis.com/documento/dc_c79be4401cf2

How to cite

Fast-VLA: You Only Think Once for Robot Manipulation — arXiv:2610.21389 — arXiv (en), https://arxiv.org/abs/2610.21389 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c79be4401cf2 (Metraipolis archive dc_c79be4401cf2, sha256 5409f9c168d8d2b548f836710aaf6c5af2e73954f9631ed44c75915f1f8cc6cb)
BibTeX
@misc{dcc79be4401cf2,
  title = {Fast-VLA: You Only Think Once for Robot Manipulation — arXiv:2610.21389},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21389},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c79be4401cf2; sha256 5409f9c168d8d2b548f836710aaf6c5af2e73954f9631ed44c75915f1f8cc6cb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21389. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-language-action (VLA) policies commonly place a multi-billion-parameter model inside the feedback loop, repeatedly recomputing task context whenever they predict a new action chunk. Yet task understanding changes far more slowly than the robot state. We introduce Fast-VLA, which invokes a frozen 4B VLM once per episode to contextualize the instruction and initial views as fixed, multi-layer key-value task memory. During execution, a lightweight online actor fuses learned time- and arm-indexed queries with current visual and proprioceptive feedback to form action queries. These queries retrieve task context from the fixed memory and predict action chunks without re-running the foundation model. Fast-VLA achieves 92.46% and 92.40% success on RoboTwin 2.0 Clean and Randomized, and 95.95% and 79.11% on LIBERO and LIBERO-Plus evaluation. Against three continually conditioned baselines on the same hardware, it reduces RoboTwin amortized policy inference time by 10.9-17.1x and requires 1.22 GiB of post-prefill online resident memory. Task-memory swaps further show that the actor relies on episode-matched task memory. These results show that task context can be established once while closed-loop control remains responsive throughout execution. Project page: this https URL

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/2610.21389 — 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 2026-10-16T08:52:17.637Z dalla pagina https://arxiv.org/list/cs.RO/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).