XiaoMaColtAI/math-modeling-skill

What it solves

This project provides a structured, three-stage workflow for AI agents to tackle mathematical modeling competitions and projects. It automates the transition from initial problem analysis and model design to code implementation and final academic paper generation, ensuring that the results are reproducible and the documentation meets professional scientific standards.

How it works

The system divides the task into three distinct roles with dedicated quality gates (checkpoints) to prevent errors from propagating:

1. Modeling Specialist: Analyzes the problem, defines the algorithm, and creates a terminology table.
2. Programming Specialist: Implements the model using Python or MATLAB, generates data tables, and produces high-quality scientific visualizations.
3. Paper Specialist: Constructs the final argument based on the results and generates the paper in Word (default) or LaTeX format.

It utilizes a "Subagent" system for optional specialized tasks (like literature review or algorithm prototyping) and a strict feedback loop where the Programming or Paper specialists can send tasks back to the Modeling specialist if implementation is impossible or evidence is missing.

Who it’s for

It is designed for users of AI agents (such as Claude Code, Cursor, Trae, or DeepSeek Harness) who are participating in mathematical modeling competitions like CUMCM, MCM/ICM, or MathorCup.

Highlights

• Dual-Language Support: Native implementation capabilities for both Python and MATLAB.
• Publication-Quality Visuals: Integrated tools for SCI/Nature-style scientific plotting with SVG and 300 DPI PNG exports.
• Dual-Engine Research: Parallel academic search using OpenAlex and AnySearch for cross-verified literature.
• Automated Paper Generation: Supports official templates with native Word OMML formulas and full LaTeX-to-DOCX conversion.
• Strict Quality Control: Implements five independent quality gates (M1, P1, P2, W1, W2) to ensure results are verified before moving to the next stage.