Mjev Spatial improves fast and accurate spatial reasoning in AI
MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
Computer Vision and Pattern Recognition
Summary
Deciding where things are or how far away they are is something humans do quickly and naturally. Existing AI models try to answer such spatial questions by generating text, which can be slow and more complex than needed. The authors created MJev-Spatial, a different approach that treats spatial tasks as classification problems, allowing much faster and simpler answers without extra model parts. They trained MJev-Spatial on many examples and tested it on various benchmarks, finding it was not only more accurate but also significantly faster than previous methods.
What this means in practice
- •For roboticists: Implement faster spatial reasoning in robots for tasks like object manipulation and navigation using MJev-Spatial’s efficient classification approach.
- •For augmented reality developers: Enable real-time, precise spatial understanding of objects and distances in AR systems without the latency of text generation.
Authors
Luozheng Qin, Yuhan Zhu, Changhai Zhou, Qian Qiao, Jun Gao, Yuhua Zhou, Harry Zhang, Xiangyu Zeng, Yuxuan Liu, Kaijie Chen, Andrew Chen, Pony Ma
Abstract
Spatial intelligence lets an agent perceive and act in the physical world, and many of its judgments must be made at interaction rates: which spatial relation holds, how far an object is, or where to place it. Vision-language models (VLMs) produce these answers by free-form text generation, paying for autoregressive decoding and answer parsing even though the output space is compact: one of a few options, a scalar, or a 2D point. In light of this, we introduce MJev-Spatial, a System 1 multimodal decision model that tackles all three types of spatial understanding tasks within a unified classification framework. Specifically, MJev-Spatial treats each candidate as a class and scores it using the backbone's pretrained query and key projections, without adding parameters. The candidates are the given options for choice tasks, or value bins and image regions refined through coarse-to-fine classification for numerical and point prediction, respectively. At inference, each classification decision takes a single forward pass and generates no answer tokens. To train on heterogeneous spatial data efficiently, we supervise all coarse-to-fine stages jointly with teacher forcing in one forward pass, and combine this with length-aware global batching and cross-GPU load balancing, raising training throughput by 30.8% over the original batching strategy. Built on Qwen3.5-4B and trained on 12.3M question-answer pairs converted from existing spatial datasets, MJev-Spatial achieves the highest overall score (65.64%) among eight tested models, ranks first on seven of twelve benchmarks, and outperforms its unmodified autoregressive base model on all twelve, while running 2.5 times faster than the fastest autoregressive baseline and 5.6 times faster than the base model.