LMST is a novel approach that uses Large Language Models to convert ML models into Solidity code that can be executed on public blockchains with optimized gas efficiency.

• Transforms both the inference path and weights of ML models into verifiable blockchain code
• Applies extensive prompt engineering to achieve gas cost optimization
• Enables on-chain verification of ML models previously limited to off-chain execution
• Bridges the gap between ML innovation and blockchain security requirements

This engineering breakthrough matters because it opens new possibilities for running verifiable AI in decentralized applications, enhancing trust and transparency in blockchain-based ML implementations.

Generation of Optimized Solidity Code for Machine Learning Models using LLMs