LLM-MM-Agent  by usail-hkust

LLM agent for complex mathematical modeling problems

Created 1 year ago
556 stars

Top 57.1% on SourcePulse

GitHubView on GitHub
Project Summary

Mathematical modeling problems are addressed by MM-Agent, an LLM-based system simulating expert human workflows. It targets students and researchers, particularly for competitions like MCM/ICM, offering an end-to-end automated solution from problem analysis to report generation, aiming to boost productivity and accuracy.

How It Works

MM-Agent simulates a four-stage human process: Problem Analysis, Mathematical Modeling, Computational Solving, and Solution Reporting. A core innovation is the Hierarchical Mathematical Modeling Library (HMML), a structured knowledge base enabling problem-aware and solution-aware retrieval of modeling strategies via an actor-critic mechanism. The Computational Solving stage autonomously generates and refines code using MLE-Solver, ensuring efficient execution.

Quick Start & Requirements

  • Primary install/run command: python MMAgent/main.py --key "your_openai_key" --task "task_id"
  • Prerequisites: Python 3.10 recommended, Conda (optional but preferred), OpenAI API key.
  • Setup: Clone the repository, create and activate a Python 3.10 Conda environment, then run pip install -r requirements.txt.
  • Links: Online Demo (Hugging Face Spaces), Paper (arXiv), Demo Video.

Highlighted Details

  • Assisted teams in achieving Finalist Awards (Top 2.0%) in MCM/ICM 2025.
  • Features an end-to-end workflow: Problem Analysis, Mathematical Modeling, Computational Solving, Solution Reporting.
  • Core capabilities include automated model selection, interactive data analysis, intelligent code generation, and professional report writing.
  • Employs a Hierarchical Mathematical Modeling Library (HMML) with 98 high-level modeling schemas.

Maintenance & Community

  • Community channels include a WeChat Group (invite via QR code in README) and email (contact@mm-agent.ai).
  • News indicates a live demo, upcoming updates, and server access prioritized by GitHub stars for MCM/ICM preparation.
  • Contributors are listed as authors in the NeurIPS 2025 paper.

Licensing & Compatibility

  • License: CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International).
  • Compatibility: Strictly non-commercial use is permitted due to the NC clause.

Limitations & Caveats

The CC BY-NC 4.0 license prohibits commercial applications. Access to beta features and server resources may be limited and prioritized based on GitHub stars. Currently, the system supports specific LLMs: OpenAI's gpt-4o and DeepSeek's deepseek-R1.

Health Check
Last Commit

5 days ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
4
Star History
41 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.