Abstract
Existing video multimodal large language models (MLLMs) organize observations primarily in time, but embodied spatial reasoning requires organizing them in space. Consequently, their visual representations can remain locally consistent across nearby frames while lacking a persistent spatial context that is stable across viewpoints and time. We exploit depth and camera pose, which are naturally available to embodied agents, to explicitly establish this spatial correspondence. We introduce Cambrian-BEV, which extends Cambrian-S with a parameter-free geometric tokenization scheme that back-projects visual patch features from each frame into a shared bird's-eye-view (BEV) coordinate frame and averages features that fall within the same spatial cell. Thus, observations of the same physical location, regardless of viewpoint or time step, are aggregated into a common spatial token. In addition to providing a persistent scene representation, this formulation acts as an efficient compression mechanism: 128 frames containing approximately 100K DINOv3 patch features are consolidated into a fixed 27$\times$27 grid of only 729 BEV tokens, cutting LLM prefill time by $2.9\times$ and peak memory by $3.6\times$. Cambrian-BEV requires no architectural modification beyond BEV tokenization and uses a frozen DINOv3 vision encoder with a trainable multimodal projector and LLM. On VSI-Bench, our model reaches 63.48, $+8.5$ points over an identically fine-tuned frame-token control with $12.6\times$ fewer visual tokens, with particularly large gains on room size ($+24.5$) and relative direction ($+34.1$). On ReVSI, it reaches 44.68, $+4.5$ over the corresponding control.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22700
