---
title: Web-CogReasoner
emoji: 🕸️
colorFrom: yellow
colorTo: indigo
sdk: gradio
sdk_version: 6.20.0
app_file: app.py
short_description: Cognitive reasoning web agent over webpage screenshots
python_version: "3.12"
startup_duration_timeout: 1h
---

# Web-CogReasoner

A demo of [Web-CogReasoner](https://huggingface.co/Gnonymous/Web-CogReasoner), a
Qwen2.5-VL-7B based multimodal web agent that performs knowledge-driven
Chain-of-Thought reasoning over webpage screenshots.

Upload a webpage screenshot, describe a task, and the model reasons step-by-step
about the page before proposing the next action.

- Paper: [arXiv:2508.01858](https://arxiv.org/abs/2508.01858)
- Code: [github.com/Gnonymous/Web-CogReasoner](https://github.com/Gnonymous/Web-CogReasoner)