Comparative Analysis of Large Language Models in Solidity Smart Contract Vulnerability Detection: Review

Abstract

Blockchain (BC) technology and Artificial Intelligence (AI) are promising areas with potential for powerful, secure, and decentralized applications in various sectors. Also, LLMs, are essential advanced machine learning frameworks, which are now used in various applications, including customer service, chatbots, code generation, and language translation. In terms of software development, LLMs are being investigated as a potential way to create smart contracts, focusing on their ability to emulate human programming techniques. For that purpose, this study reviews the capabilities and limitations of Large Language Models (LLMs) like GPT-3.5, PaLM2, and Code Llama in detecting vulnerabilities in Solidity smart contract code.