---
title: CodeFuse-SVR-8B
emoji: 🔍
colorFrom: green
colorTo: red
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
short_description: Structured visual reasoning for screenshots and diagrams
python_version: "3.12"
startup_duration_timeout: 30m
---

# CodeFuse-SVR-8B: Structured Visual Reasoning

This Space demonstrates **CodeFuse-SVR-8B**, a vision-language model fine-tuned
from Qwen3-VL for the [SVRepair](https://arxiv.org/abs/2602.06090) framework.

The model classifies an input image into one of six categories and generates a
structured representation:

| Category | Output |
|---|---|
| Mermaid Diagram | Mermaid code block |
| Data Visualization | Python (matplotlib) code |
| Webpage Screenshot | HTML structure |
| Screenshot or Document | Extracted text |
| Natural Image | Natural-language description |
| Other Image | High-level summary |

Upload a GUI screenshot, chart, diagram, or photo to see the model's structured
visual reasoning output.