Scholarlink

<p>Custom generative pre-trained transformers (GPTs) are transforming domain-specific problem-solving across scientific fields. However, materials science lacks a dedicated, accessible GPT-based assistant tailored to its unique challenges. To address this gap, I present ChatGPT Material Explorer 1.0 (CME), a custom GPT that integrates large language models with graph neural networks (GNN) models and other domain-specific APIs to enhance materials design workflows. The assistant offers core functionalities such as: (1) intelligent exploration of molecular and materials databases, (2) GNN-driven prediction of materials properties, (3) efficiently searching arXiv for papers, and (4) interactive, conversational support for scientific analysis and writing. I evaluate its performance against general-purpose models such as OpenAI GPT-4o highlighting its superior ability to retrieve accurate data, reduce hallucination, and provide context-aware responses. By combining natural language understanding with real-time data access and physics-informed modeling, CME aims to democratize computational tools and accelerate decision-making in materials research. More details about CME are available at: <a href="https://github.com/AtomGPTLab/chatgpt_material_explorer">https://github.com/AtomGPTLab/chatgpt_material_explorer</a> to facilitate broader community adoption.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ChatGPT Material Explorer: Design and Implementation of a Custom GPT Assistant for Materials Science Applications

  • 期刊论文

  • 发表日期：

    2025年7月9日

查看全文
  • Kamal Choudhary

摘要

Custom generative pre-trained transformers (GPTs) are transforming domain-specific problem-solving across scientific fields. However, materials science lacks a dedicated, accessible GPT-based assistant tailored to its unique challenges. To address this gap, I present ChatGPT Material Explorer 1.0 (CME), a custom GPT that integrates large language models with graph neural networks (GNN) models and other domain-specific APIs to enhance materials design workflows. The assistant offers core functionalities such as: (1) intelligent exploration of molecular and materials databases, (2) GNN-driven prediction of materials properties, (3) efficiently searching arXiv for papers, and (4) interactive, conversational support for scientific analysis and writing. I evaluate its performance against general-purpose models such as OpenAI GPT-4o highlighting its superior ability to retrieve accurate data, reduce hallucination, and provide context-aware responses. By combining natural language understanding with real-time data access and physics-informed modeling, CME aims to democratize computational tools and accelerate decision-making in materials research. More details about CME are available at: https://github.com/AtomGPTLab/chatgpt_material_explorer to facilitate broader community adoption.