Metraipolis › archive › dc_6c29d3f3ff40

VLA-$Δ$: Learning Physical State Transition for Vision-Language-Action Models — arXiv:2610.21306

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21306
License (as archived)
CC0-1.0
Archived
2026-10-16T10:35:15.047Z
SHA-256 of the archived text
5f979adca7f2c4a765f53e193fa6fc6b25a0593c9a92e48c09b82da96b910528
Stable URL of this page
https://metraipolis.com/documento/dc_6c29d3f3ff40

How to cite

VLA-$Δ$: Learning Physical State Transition for Vision-Language-Action Models — arXiv:2610.21306 — arXiv (en), https://arxiv.org/abs/2610.21306 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6c29d3f3ff40 (Metraipolis archive dc_6c29d3f3ff40, sha256 5f979adca7f2c4a765f53e193fa6fc6b25a0593c9a92e48c09b82da96b910528)
BibTeX
@misc{dc6c29d3f3ff40,
  title = {VLA-$Δ$: Learning Physical State Transition for Vision-Language-Action Models — arXiv:2610.21306},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21306},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6c29d3f3ff40; sha256 5f979adca7f2c4a765f53e193fa6fc6b25a0593c9a92e48c09b82da96b910528},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21306. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-language-action (VLA) models transfer semantic knowledge acquired through vision-language pretraining to robot control, achieving strong performance across various benchmarks. However, most VLAs directly map observations to demonstrated actions without explicit supervision of intermediate physical state changes, making them prone to overfitting observation-action correspondences. Some methods introduce intermediate reasoning or future-state prediction in language, visual, or latent spaces to guide action generation, but do not explicitly supervise the physical state transition itself. We therefore propose VLA-$\Delta$, a general framework that uses physical changes between current and future states as explicit learning targets. To support this objective, we construct a large-scale pretraining dataset and derive physical state-transition annotations covering task-relevant semantics and visual motion changes, providing more direct supervision for continuous action generation. During inference, the model generates actions directly from current observations without additional inference overhead. VLA-$\Delta$ achieves performance competitive with the state of the art on LIBERO and RoboTwin, and outperforms $\pi_{0.5}$ in real-robot experiments. Our code is available at this https URL

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.21306 — 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 2026-10-16T08:52:02.650Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).