Urbex turns city-scale 3D grounding into active spatial search, tops CityRefer
alibabagroup · hf · 2026-10-14
Urbex (Wuhan University & Alibaba DreamX) recasts city-scale 3D grounding from candidate ranking into active spatial search: a VLM agent interacts with the scene as a multi-view environment — querying landmarks, zooming in, rendering oblique views, and committing a bounding box — with the tool-use policy optimized via RL on CityRefer using smooth localization and shaping rewards. Urbex achieves the best Hit@1 and IoU on CityRefer, supports candidate-free grounding, is substantially faster than the strongest ranking baseline, and generalizes zero-shot on STPLS3D-Refer. Code is open-sourced.
More from Research
- RUC and Kling AI propose EventLens: event-structure RL to fix video LLMs' temporal understanding — Guowen Zhang · 2026-10-14
- JSIR dataset drops: 267 jazz recordings with beat, chord and structure annotations — umpedronosapato · 2026-10-14
- There is no such thing as an AI drug, argues researcher Patrick Kidger — PatrickKidger · 2026-10-14
- More thinking, worse scores: refusals tank closed-model evals at max reasoning effort — _lewtun · 2026-10-14
- Integer multiplication repo becomes an agent harness benchmark — and collaboration loses — iskander · 2026-10-14
- Recurrent self-conditioning: framing recent diffusion methods as depth decoupled from time — alec_helbling · 2026-10-14