Huixin Xu, Ziya Yu, Peng Li · 2026 · arXiv
DOI: 10.48550/arXiv.2610.31423 · open-access PDF
Los modelos de visión-lenguaje-acción (VLA) se han convertido en un paradigma importante para la manipulación robótica al unificar la percepción visual, la comprensión del lenguaje y la generación de acción. Dado que la manipulación confiable requiere una observación suficiente de objetos, efectos finales y sus relaciones espaciales, las entradas multi-visión se utilizan ampliamente para proporcionar información espacial complementaria. Sin embargo, nuestros experimentos muestran que los modelos VLA existentes no aprovechan plenamente esta complementariedad multi-visión.
Machine translation (Meta M2M100); the English original follows.
Vision-language-action (VLA) models have become an important paradigm for robotic manipulation by unifying visual perception, language understanding, and action generation. Since reliable manipulation requires sufficient observation of objects, end-effectors, and their spatial relationships, multi-view inputs are widely used to provide complementary spatial information. However, our experiments show that existing VLA models do not fully exploit such multi-view complementarity. The policy tends to persistently rely on the wrist view, lacking the ability to dynamically adjust global-local view d
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=2610.31423, 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