---
title: SpatioLM Spatial Understanding
emoji: 🧭
colorFrom: gray
colorTo: yellow
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
short_description: Spatial reasoning VLM demo powered by SpatioLM
python_version: "3.12"
startup_duration_timeout: 30m
---

# SpatioLM: Spatial Understanding Demo

This Space demonstrates **SpatioLM-Understanding-InternVL3.5**, a vision-language
model with enhanced physical spatial intelligence. Upload an image and ask
spatial-reasoning questions such as "Which object is closer to the camera?" or
"Describe the spatial arrangement of objects in this scene."

SpatioLM introduces a plug-and-play spatio-vision module that learns physically
coherent representations from pseudo depth and camera supervision — without
requiring 3D input at inference time.

## Links

- [Paper (arXiv)](https://arxiv.org/abs/2608.01899)
- [GitHub](https://github.com/xiaomi-research/spatio-lm)
- [Model on Hugging Face](https://huggingface.co/xiaomi-research/SpatioLM-Understanding-InternVL3.5)