Metraipolis › archive › dc_bb47b2eed93f

REKI: Registering Knowledge in VLAs — arXiv:2610.22178

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.22178
License (as archived): CC0-1.0
Archived: 2026-10-16T10:25:22.174Z
SHA-256 of the archived text: 86d98202fd9b318c342f8b6057ee1dbf1d44738e632a8314f5049f82e4fd3e4d
Stable URL of this page: https://metraipolis.com/documento/dc_bb47b2eed93f

How to cite

REKI: Registering Knowledge in VLAs — arXiv:2610.22178 — arXiv (en), https://arxiv.org/abs/2610.22178 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bb47b2eed93f (Metraipolis archive dc_bb47b2eed93f, sha256 86d98202fd9b318c342f8b6057ee1dbf1d44738e632a8314f5049f82e4fd3e4d)

BibTeX

@misc{dcbb47b2eed93f,
  title = {REKI: Registering Knowledge in VLAs — arXiv:2610.22178},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.22178},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bb47b2eed93f; sha256 86d98202fd9b318c342f8b6057ee1dbf1d44738e632a8314f5049f82e4fd3e4d},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22178. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We introduce REKI, a 2.4B vision-language-action model (VLA) pretrained exclusively on public data that sets a new Pareto frontier between performance and model size: on LIBERO-Plus, RoboTwin Clean2Rand and RoboCasa365, it is outperformed only by models about twice its size. REKI consists of a vision-language model (VLM) followed by a flow-matching action expert, and is pretrained on a curated mixture of public robot, egocentric human and vision-language data. Additionally, we endow it with a memory that gives the policy access to its recent past at negligible cost: a few register tokens in the action expert summarise the output of the VLM at each frame and persist in the expert's cache. This memory improves success in simulation and on real robots, and makes oscillations between poses about four times rarer. REKI uses a simple model design and training paradigm, and is optimised to be efficient at training and inference. We release the full code across data curation, training, inference and evaluation, deployment, and all pretrained and finetuned model weights: this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.22178 — 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 2026-10-16T08:48:02.940Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).