@inproceedings{karanjai2024solmoverfeasibility,
title={SolMover: Feasibility of Using LLMs for Translating Smart Contracts},
author={Karanjai, Rabimba and Xu, Lei and Shi, Weidong},
booktitle={2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC)},
pages={1--3},
year={2024}
}Copied to clipboardFull paper available on the publisher's website.
Open via DOI →We investigate the feasibility of employing large language models (LLMs) to translate smart contracts, focusing on their ability to emulate human translation techniques. This inquiry centers on the potential of LLMs to comprehend a smart contract written in Solidity and to produce an equivalent in Move, a low-resource language. We introduce SolMover, a system designed to bridge the gap between Solidity and Move through a multi-step knowledge distillation process. Our study aims to elucidate whether LLMs can internalize coding concepts, generate specific sub-tasks from a general prompt, produce compilable code in a language they were not extensively trained on, and how compiler feedback can be used to debug and enhance the quality of the generated code.