Search papers, labs, and topics across Lattice.
The paper introduces SpatiO, a heterogeneous multi-agent framework for spatial reasoning that leverages vision-language specialists with complementary inductive biases like 2D appearance, depth, and geometric constraints. To enable effective collaboration, they propose Test-Time Orchestration (TTO), an optimization mechanism that dynamically reweights agents based on their observed reliability during inference, without requiring parameter updates. Experiments on spatial reasoning benchmarks show SpatiO consistently outperforms existing baselines by effectively adapting to varying contextual cues.
Spatial reasoning gets a boost: a new framework dynamically orchestrates vision-language agents at test time, outperforming fixed-pipeline approaches by adapting to the reliability of different spatial cues.
Understanding visual scenes requires not only recognizing objects but also reasoning about their spatial relationships. Unlike general vision-language tasks, spatial reasoning requires integrating multiple inductive biases, such as 2D appearance cues, depth signals, and geometric constraints, whose reliability varies across contexts. This suggests that effective spatial reasoning requires \emph{spatial adaptability}: the ability to flexibly coordinate different reasoning strategies depending on the input. However, most existing approaches rely on a single reasoning pipeline that implicitly learns a fixed spatial prior, limiting their ability to adapt under distribution changes. Multi-agent systems offer a promising alternative by aggregating diverse reasoning trajectories, but prior attempts in spatial reasoning primarily employ homogeneous agents, restricting the diversity of inductive biases they can leverage. In this work, we introduce \textbf{\textsc{SpatiO}}, a heterogeneous multi-agent framework for spatial reasoning that coordinates multiple vision-language specialists with complementary inductive biases. To enable effective collaboration, we propose \textbf{Test-Time Orchestration (TTO)}, an optimization mechanism that dynamically evaluates and reweights agents based on their observed reliability during inference, without modifying model parameters. Extensive experiments on diverse spatial reasoning benchmarks, including 3DSRBench, STVQA-7k, CV-Bench, and Omni3D-Bench, demonstrate that \textsc{SpatiO} consistently improves spatial reasoning performance over both closed-source and open-source baselines.
