Application of Self-Evolving AI Agents in Chemical Research: A Novel Intelligent Assistance System 🧪🤖

🧬 Paper | 🔬 Blog Post | 📊 Data

This project involves fine-tuning open-source large language models with chemical science data, evaluated using a specialized automated scoring system. The resulting chemical intelligent assistant system utilizes the fine-tuned Mistral Nemo model and can flexibly incorporate various advanced models. It integrates chemistry-specific features like molecular visualization and literature retrieval, while also possessing autonomous evolution capabilities through knowledge accumulation, skill acquisition, and collaborative mechanisms. This approach enables continuous optimization of the system's professional abilities and interaction quality, overcoming limitations of traditional static AI systems in the chemistry domain.

🌟 Key Features

Fine-tuning Large Language Models Based on Chemistry Domain Data

Utilizing collected and curated chemistry instruction data
Fine-tuning mainstream open-source large language models
Developing a specialized automatic scoring system for the chemistry domain

Innovative Chemical Intelligent Assistant System Design

Using the fine-tuned Mistral Nemo model as one of the primary models
Incorporating mechanisms for flexible invocation of various advanced models
Continuously leveraging the latest AI capabilities, considering the rapid iteration of large language models

Deep Integration of Chemistry Expertise and Requirements

Integrating professional functions such as molecular visualization, SMILES string processing, and chemical literature retrieval
Significantly enhancing the system's practical value in chemical research and applications

Autonomous Evolution Capability

Through knowledge accumulation, skill acquisition, performance evaluation, and collective collaboration mechanisms
Continuously optimizing professional capabilities and interaction quality
Breaking through the inherent static limitations of traditional AI systems

📋 Requirements

Installation

# Create and activate a new conda environment
conda create -n gvim python=3.9.19
conda activate gvim

# Install required packages
pip install -r requirements.txt

citation

{ "name": "Kangyong Ma", "affiliation": { "institution": "College of Physics and Electronic Information Engineering, Zhejiang Normal University", "city": "Jinhua City", "postalCode": "321000", "country": "China" "DOI":"10.26434/chemrxiv-2024-6tv8c" }, "email": [ "[email protected]", "[email protected]" ] }