Metraipolis › archive › dc_bc4cd6523103

Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies — arXiv:2609.24682

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.24682
License (as archived)
CC0-1.0
Archived
2026-10-09T22:06:40.698Z
SHA-256 of the archived text
097dd49007b1d772e3cafb4ae72605f606ca2831c3b4629ed2bcc88651539fea
Stable URL of this page
https://metraipolis.com/documento/dc_bc4cd6523103

How to cite

Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies — arXiv:2609.24682 — arXiv (en), https://arxiv.org/abs/2609.24682 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bc4cd6523103 (Metraipolis archive dc_bc4cd6523103, sha256 097dd49007b1d772e3cafb4ae72605f606ca2831c3b4629ed2bcc88651539fea)
BibTeX
@misc{dcbc4cd6523103,
  title = {Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies — arXiv:2609.24682},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.24682},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bc4cd6523103; sha256 097dd49007b1d772e3cafb4ae72605f606ca2831c3b4629ed2bcc88651539fea},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.24682. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-Language-Action (VLA) models map observations to actions with no objective that accounts for how the world responds, so their robustness is bounded primarily by data coverage. World models carry precisely that missing objective and are better grounded for it, yet rolling the future forward costs seconds per decision and rules them out of the control loop. We show the two can be separated. What a world model knows about physical scenes lives in its internal features; generating the future is merely the objective that produced them, so the grounding can be inherited while the generative machinery is left behind. We add one feature-alignment term to ordinary VLA training: a frozen world model is run over the training frames once and cached, and the student learns to agree with that cache. No teacher is loaded during training, the projector is discarded after it, and the deployed policy is identical to the undistilled baseline, running in 32ms and 1.86GB on a consumer RTX5090, so every gain is attributable to the representation rather than to added capacity or test-time compute. A 0.8B student reaches 97.9% on LIBERO, improves from 48.2% to 50.5% on RoboCasa-GR1 humanoid manipulation, and the same objective carries over to real hardware, on both a single-arm and a bimanual platform. The gain survives changes of student scale, backbone, alignment layer, and teacher, indicating a broad representational prior rather than a fragile alignment between two particular networks. Project page: this https URL .

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CV).
Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.24682 — 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 2026-10-09T21:45:21.156Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).