AI RESEARCH
UniT: Unified Geometry Learning with Group Autoregressive Transformer
arXiv CS.CV
•
ArXi:2605.21131v1 Announce Type: new Recent feed-forward models have significantly advanced geometry perception for inferring dense 3D structure from sensor observations. However, its essential capabilities remain fragmented across multiple incompatible paradigms, including online perception, offline reconstruction, multi-modal integration, long-horizon scalability, and metric-scale estimation. We present UniT, a unified model built upon a novel Group Autoregressive Transformer, which reformulates these seemingly disparate capabilities within a single framework.