Firehose

30 JUL 2026 · Paper

SpatialCLI: Learning to Reason With Spatial Tools, Then Without Them

This paper develops a framework, SpatialCLI, to help vision-language models (VLMs) better understand and use visual tools to make better decisions. By training VLMs to reason with spatial tools and then internalize those capabilities, SpatialCLI can improve the performance of VLMs in tasks that require visual reasoning.

vision-language models spatial reasoning visual perception embodied agents reinforcement learning

Browse by tag

SpatialCLI: Learning to Reason With Spatial Tools, Then Without Them