Rui Gong, Ruoyu Wang, Yalun Dai, Hao Li, Sidi Yang, Qihao Zhao, Hongyu Li, Zeyang Guan · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28354 · open-access PDF
O raciocínio espacial exige que um modelo descubra onde os objetos estão no espaço 3D, como eles se relacionam uns com os outros e como eles se movem. VLMs agenciais recentemente trouxeram grandes melhorias em tais tarefas, permitindo que o modelo recolha ativamente evidências com ferramentas de percepção 3D em vez de responder em uma única passagem. Mesmo assim, em muitas perguntas, as ferramentas devolvem a evidência correta e o modelo ainda não chega à resposta correta: a lacuna entre o modelo e as ferramentas, onde as leituras brutas se tornam conclusões usáveis, é deixada sem preencher.
Machine translation (Meta M2M100); the English original follows.
Spatial reasoning asks a model to work out where objects are in 3D space, how they relate to one another, and how they move. Agentic VLMs have recently brought large improvements on such tasks by letting the model actively gather evidence with 3D perception tools instead of answering in a single pass. Even so, on many questions the tools return the right evidence and the model still does not arrive at the correct answer: the gap between model and tools, where raw readings become usable conclusions, is left unfilled. We present S-Agent 2.0, a spatial reasoning agent that equips an off-the-shelf
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=2610.28354, 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