Search papers, labs, and topics across Lattice.
This paper introduces SpatialCLI, a framework designed to enhance vision-language models (VLMs) by enabling them to reason with spatial tools and subsequently internalize the perceptual capabilities of these tools. The framework operates in three stages: it first augments VLM perception with specialist vision models, then employs Cold-Start SFT and agentic reinforcement learning to refine tool usage, and finally verbalizes successful trajectories to internalize these capabilities. The results show a significant performance improvement on the MindCube benchmark, with the Qwen3-VL-8B-Instruct model achieving an accuracy of 84.6% when using tools, outperforming a leading model while maintaining strong performance without tools after internalization.
SpatialCLI enables VLMs to achieve a staggering 84.6% accuracy on spatial reasoning tasks, far surpassing existing models.
Vision-language models (VLMs) are increasingly used in embodied agents to interpret visual inputs, reason about spatial relationships, and make task-level decisions based on that reasoning. However, a fundamental capability mismatch remains: general VLMs can reason about the overall task but often miss the visual details that determine success, while specialist vision models can capture those details but cannot translate them into task-level decisions. In this work, we propose SpatialCLI, a framework that teaches VLMs to reason with spatial tools and progressively internalize the specialist perceptual capabilities they provide. SpatialCLI proceeds in three stages: (1) Call exposes specialist vision models as spatial tools to augment the VLM's perception; (2) Learn uses Cold-Start SFT and agentic RL to improve tool use; and (3) Internalize verbalizes successful tool-use trajectories to internalize specialist perceptual capabilities. We further introduce SpatialCLI-Bench, a 516-example benchmark for compositional perception across localization, segmentation, depth, and pose. On MindCube, SpatialCLI raises Qwen3-VL-8B-Instruct from 29.3% to 84.6% with tools, surpassing GPT-5.6 Sol with tools (72.1%), while retaining 73.8% without tools after internalization.
