LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding

Shihao Wang, Shilong Liu, Yuanguo Kuang, Xinyu Wei, Yangzhou Liu, Zhiqi Li, Yunze Man, Guo Chen, Andrew Tao, Guilin Liu, Jan Kautz, Lei Zhang, Zhiding Yu

arXiv:2605.27365 · 2026-05-27 공개 · arXiv · PDF

vlm large-scale-training localization unified-grounding parallel-box-decoding box-decoding high-precision-localization

Abstract

Vision-language models (VLMs) commonly formulate visual grounding and detection as a coordinate-token generation problem, serializing each 2D box into multiple 1D tokens that are learned and decoded largely independently. This token-by-token decoding mismatches the coupled structure of box geometry and creates a practical inference bottleneck due to strictly sequential generation. We introduce LocateAnything, a unified generative grounding and detection framework based on Parallel Box Decoding (PBD). By decoding geometric elements such as bounding boxes and points as atomic units in a single step, LocateAnything preserves intra-box geometric coherence and unlocks substantial parallelism. We show that PBD improves both decoding throughput and localization accuracy. We further develop a scalable data engine and curate LocateAnything-Data, a large-scale dataset with more than 138 million training samples, substantially increasing data diversity for high-precision localization. Extensive evaluations show that LocateAnything advances the speed-accuracy frontier, achieving significantly higher decoding throughput while improving high-IoU localization quality across diverse benchmarks. The results highlight the complementary benefits of Parallel Box Decoding and large-scale training data in enabling efficient and precise unified visual grounding and detection.

한국어 요약

📋 한 줄 요약

**[Visual Grounding / Parallel Decoding]** LocateAnything가 Parallel Box Decoding(PBD)으로 bounding box를 single step에 atomic하게 디코딩 — sequential 토큰 생성 병목 해소, 138M+ 샘플 데이터셋과 결합해 throughput·high-IoU localization 동시 frontier push.

🎯 핵심 기여도

💡 핵심 아이디어

VLM의 grounding은 box를 토큰 시퀀스로 직렬화하지 말고 atomic geometric unit으로 한 번에 parallel decoding해야 한다 — 이는 intra-box geometric coherence를 보존하고 sequential 병목을 제거하면서 138M+ 대규모 데이터와 결합 시 speed·accuracy frontier를 동시 push한다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: VLM grounding의 token serialization 한계를 atomic decoding으로 해소, throughput·accuracy 동시 개선이라는 보기 드문 trade-off 극복, 138M+ 데이터로 high-precision localization 새 기준, generative grounding·detection 통합으로 generality. **한계**: 138M+ 데이터 큐레이션·학습 비용, PBD가 매우 복잡한 다 객체·관계 추론에 미적용, atomic unit 정의의 다른 기하 task(polygon·mask) 확장 여지.

🚀 실용적 활용