MathAgent: Leveraging a Mixture-of-Math-Agent Framework for Real-World Multimodal Mathematical Error Detection

TL;DR

MathAgent tackles the problem of multimodal mathematical error detection in real-world education by decomposing the task into three specialized agents: Image-Text Consistency Validator, Visual Semantic Interpreter, and Integrative Error Analyzer. It demonstrates about 5% gain in error-step identification and 3% gain in error categorization on real data, and has been deployed on platforms serving over a million K-12 students with high satisfaction and cost savings. The approach explicitly models cross-modal relationships between problems, visuals, and student reasoning to improve accuracy and scalability beyond prior MLLM-based methods. The work highlights practical educational impact and provides a framework for integrating multimodal reasoning into scalable feedback systems.

Abstract

Mathematical error detection in educational settings presents a significant challenge for Multimodal Large Language Models (MLLMs), requiring a sophisticated understanding of both visual and textual mathematical content along with complex reasoning capabilities. Though effective in mathematical problem-solving, MLLMs often struggle with the nuanced task of identifying and categorizing student errors in multimodal mathematical contexts. Therefore, we introduce MathAgent, a novel Mixture-of-Math-Agent framework designed specifically to address these challenges. Our approach decomposes error detection into three phases, each handled by a specialized agent: an image-text consistency validator, a visual semantic interpreter, and an integrative error analyzer. This architecture enables more accurate processing of mathematical content by explicitly modeling relationships between multimodal problems and student solution steps. We evaluate MathAgent on real-world educational data, demonstrating approximately 5% higher accuracy in error step identification and 3% improvement in error categorization compared to baseline models. Besides, MathAgent has been successfully deployed in an educational platform that has served over one million K-12 students, achieving nearly 90% student satisfaction while generating significant cost savings by reducing manual error detection.

Figures (4)

• Figure 1: Comparison between previous human-based (a) and MLLM-based (b) paradigms vs. our proposed MathAgent framework (c) for multimodal mathematical error detection.
• Figure 2: The framework of our proposed Mixture-of-Math-Agent for multimodal mathematical error detection.
• Figure 3: Ablation study of MathAgent.
• Figure 4: Key statistics of dataset.