LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding
Abstract
Parallel Box Decoding enables efficient and accurate unified visual grounding and detection by decoding geometric elements as atomic units, improving both throughput and localization quality.
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.
Community
the block-level encoding with a fixed length L=6 that packs a bounding box and two structural tokens is a surprisingly crisp way to align geometry with parallel decoding. i like that they maintain two aligned streams, a standard next-token prediction and a block-level multi-token prediction, with a specialized attention mask that keeps them isolated yet shares context. btw arxivlens has a solid breakdown that helped me parse the details here: https://arxivlens.com/PaperView/Details/locateanything-fast-and-high-quality-vision-language-grounding-with-parallel-box-decoding-5024-c610b253. one practical question i have is how sensitive the performance is to the chosen block length or to padding in really crowded scenes where many boxes compactly overlap? overall, the data scale in locateanything-data plus this parallel decoding seems to be the right recipe to push both speed and high-iou accuracy in real-world grounding tasks.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Fast-dVLM: Efficient Block-Diffusion VLM via Direct Conversion from Autoregressive VLM (2026)
- Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects (2026)
- Reasoning-Guided Grounding: Elevating Video Anomaly Detection through Multimodal Large Language Models (2026)
- Qwen3-VL-Seg: Unlocking Open-World Referring Segmentation with Vision-Language Grounding (2026)
- Orthrus: Memory-Efficient Parallel Token Generation via Dual-View Diffusion (2026)
- AnchorSeg: Language Grounded Query Banks for Reasoning Segmentation (2026)
- Small Vision-Language Models are Smart Compressors for Long Video Understanding (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2605.27365 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 1
Datasets citing this paper 0
No dataset linking this paper