PeijieWang/GDP-4B

VISIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Apr 23, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

PeijieWang/GDP-4B is a 4 billion parameter vision-language model developed by Peijie Wang and collaborators, specifically designed for geoparsing. This model excels at parsing geometric diagrams, including both plane and solid geometry, and translating them into a unified formal language. It is built upon the Qwen3-VL architecture and is optimized for understanding and describing complex geometric figures from images.

Loading preview...

Overview

PeijieWang/GDP-4B is a 4 billion parameter vision-language model (VLM) developed by Peijie Wang et al., specialized in geoparsing. This model is designed to interpret and formally describe geometric diagrams, encompassing both 2D plane geometry and 3D solid geometry. It leverages a unified formal language for its output, making it suitable for automated geometric reasoning tasks.

Key Capabilities

  • Geometric Diagram Parsing: Accurately analyzes images of geometric diagrams.
  • Formal Language Description: Translates visual geometric information into a structured, unified formal language.
  • Plane and Solid Geometry: Handles both two-dimensional and three-dimensional geometric figures.
  • Qwen3-VL Architecture: Built upon the Qwen3-VL framework, indicating strong multimodal processing capabilities.

Good For

  • Automated Geometry Problem Solving: Generating formal representations from diagrams for solvers.
  • Educational Tools: Assisting in the understanding and formalization of geometric concepts.
  • Research in AI for Mathematics: Providing a foundation for further development in automated geometric reasoning.

For more details, refer to the project homepage and the associated research paper.

Finetunes
1 models