< Explain other AI papers

TrustGeoGen: Scalable and Formal-Verified Data Engine for Trustworthy Multi-modal Geometric Problem Solving

Daocheng Fu, Zijun Chen, Renqiu Xia, Qi Liu, Yuan Feng, Hongbin Zhou, Renrui Zhang, Shiyang Feng, Peng Gao, Junchi Yan, Botian Shi, Bo Zhang, Yu Qiao

2025-04-29

TrustGeoGen: Scalable and Formal-Verified Data Engine for Trustworthy
  Multi-modal Geometric Problem Solving

Summary

This paper talks about TrustGeoGen, a new system that creates and checks a wide variety of geometry problems using both pictures and words, making sure the problems and answers are correct through formal logic.

What's the problem?

The problem is that current AI systems for solving geometry problems often struggle with accuracy and can't handle new or unusual types of questions very well, partly because the data they train on isn't always reliable or varied enough.

What's the solution?

The researchers built TrustGeoGen to automatically generate lots of different geometry problems that include both text and images, and then use formal logic to double-check that the answers are correct. This creates a trustworthy set of practice problems and solutions that AI models can use to learn and improve.

Why it matters?

This matters because it helps AI become much better at solving geometry problems, even ones it hasn't seen before, which is useful for education, tutoring, and any field that relies on understanding shapes and spaces.

Abstract

TrustGeoGen is a scalable data engine providing a formally verified benchmark for GPS through innovative multimodal generation and logical verification, improving model accuracy and OOD generalization.