Cambrian-P: Pose-Grounded Video Understanding

TL;DR

Cambrian-P introduces camera pose as a supervisory signal in multimodal large language models for video understanding, significantly improving spatial reasoning and generalization across multiple benchmarks.

Contribution

The paper proposes Cambrian-P, a novel video MLLM that incorporates learnable camera tokens and pose regression, demonstrating the importance of pose information in video reasoning tasks.

Findings

Achieves 4.5-6.5% gains on VSI-Bench spatial reasoning benchmark.

Generalizes well across eight additional spatial and video QA benchmarks.

State-of-the-art streaming pose estimation on ScanNet.

Abstract

Camera pose matters. The position and orientation of each viewpoint define a shared spatial coordinate frame that relates observations across video frames. Yet this signal is largely absent from multimodal LLMs (MLLMs) for video understanding, which process frames as isolated 2D snapshots, instead of the persistent scene humans perceive. We revisit pose as a lightweight supervisory signal and introduce Cambrian-P, a video MLLM augmented with per-frame learnable camera tokens and a pose regression head. With a carefully designed sampling scheme, the model achieves substantial gains of 4.5-6.5% on spatial reasoning benchmarks such as VSI-Bench, generalizes across eight additional spatial and general video QA benchmarks, and, as a byproduct, achieves state of the art streaming pose estimation on ScanNet. Surprisingly, training on pseudo-annotated poses from in-the-wild video further improves general video QA benchmarks, showing pose helps beyond spatial reasoning. Together, these results position camera pose as a fundamental signal for video models that reason about the physical world.