Yunfei Xie, Howard H. Qian, Shiyi Lan, Kaiyu Hang, Chen Wei · 2026 · arXiv
DOI: 10.48550/arXiv.2610.31670 · open-access PDF
Los modelos de fundación robótica suelen ser evaluados a través de costosos lanzamientos de ciclo cerrado, lo que motiva a los proxies offline baratos para la capacidad de modelar. La hipótesis de representación platónica (PRH) sugiere que a medida que las representaciones mejoran, diferentes modalidades como la visión y el lenguaje convergen hacia un modelo estadístico compartido de la realidad. Nos preguntamos si las trayectorias de movimiento forman otra modalidad similar: a medida que los modelos encarnados se vuelven más capaces, ¿deberían sus representaciones sensoriales converger hacia la estructura del movimiento físico?
Machine translation (Meta M2M100); the English original follows.
Robotic foundation models are typically evaluated through costly closed-loop rollouts, which motivates cheap offline proxies for model capability. The Platonic Representation Hypothesis (PRH) suggests that as representations improve, different modalities such as vision and language converge toward a shared statistical model of reality. We ask whether motion trajectories form another such modality: as embodied models become more capable, should their sensory representations converge toward the structure of physical motion? To examine this, we propose Sensorimotor Alignment ($S^2$), a training-f
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31670, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange