DeepYardDeepYard
S

SpatialCLI

Framework teaching vision-language models spatial reasoning through tool use and internalization

Open SourceFree

About

SpatialCLI bridges general vision-language model capabilities with specialist spatial understanding for embodied AI agents. The framework teaches VLMs to reason using external spatial tools, then internalize those reasoning patterns for improved geometric and spatial tasks. Designed for robotics and embodied AI applications requiring sophisticated spatial comprehension beyond standard VLM capabilities.

Details

Language
Patterns

Tags

frameworktool-useautonomousopen-sourcepython